I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
My server, an HP MicroServer, came back (re-booted) on 2 of the 3 occasions, but not on the third.
I assume that the problem arises because the machine does not close down properly. (Although it is also possible that a voltage surge might have been responsible - I have no surge protector on this supply.)
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
I know there is a Remote Management (iLO) card for this machine, which might be useful for this. Unfortunately, I've already used the PCIe slot for a second ethernet card.
Any advice or suggestions gratefully received.
On Fri, Jul 1, 2011 at 7:05 PM, Timothy Murphy gayleard@eircom.net wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
I believe a personal UPS would be quite cheap. Much cheaper than losing data. The UPS can tell the OS to shutdown on power loss.
On Fri, 1 Jul 2011, Timothy Murphy wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
<snip>
Just buy a really basic UPS. I don't know what the prices are like where you are, but a crappy 500VA UPS can be had for about 25 uk pounds.
I've only ever monitored APC UPSs which can be monitored easily from linux, so check for linux compatibility before buying something obscure.
jh
At Fri, 1 Jul 2011 12:26:10 +0100 (BST) CentOS mailing list centos@centos.org wrote:
On Fri, 1 Jul 2011, Timothy Murphy wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
<snip>
Just buy a really basic UPS. I don't know what the prices are like where you are, but a crappy 500VA UPS can be had for about 25 uk pounds.
I've only ever monitored APC UPSs which can be monitored easily from linux, so check for linux compatibility before buying something obscure.
With a non-Linux compatable UPS, you can use a old analog serial modem as a power sensor. If the machine has a serial port (RS-232), you can plug the modem into the wall outlet and connect it to the computer's serial port. When the power goes out, the modem goes off and powerd can sense the loss of Modem Ready and treat that as a 'power failure' signal. This trick works for cheap, obscure or basic *dumb* UPSs.
jh _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, 1 Jul 2011, Robert Heller wrote:
With a non-Linux compatable UPS, you can use a old analog serial modem as a power sensor. If the machine has a serial port (RS-232), you can plug the modem into the wall outlet and connect it to the computer's serial port. When the power goes out, the modem goes off and powerd can sense the loss of Modem Ready and treat that as a 'power failure' signal. This trick works for cheap, obscure or basic *dumb* UPSs.
Nice. ;)
Another trick I used was to hook the monitor up to non-UPS power, and connect the USB hub within it to the PC. A udev trigger than runs a script when the device appears or disappers. In my case it was to reconfigure the displays when monitors were turned on, but I'd not thought of using it for UPS monitoring.
jh
Robert Heller wrote:
At Fri, 1 Jul 2011 12:26:10 +0100 (BST) CentOS mailing list centos@centos.org wrote:
On Fri, 1 Jul 2011, Timothy Murphy wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
<snip>
Just buy a really basic UPS. I don't know what the prices are like where you are, but a crappy 500VA UPS can be had for about 25 uk pounds.
I've only ever monitored APC UPSs which can be monitored easily from linux, so check for linux compatibility before buying something obscure.
With a non-Linux compatable UPS, you can use a old analog serial modem as a power sensor. If the machine has a serial port (RS-232), you can plug the modem into the wall outlet and connect it to the computer's serial port. When the power goes out, the modem goes off and powerd can sense the loss of Modem Ready and treat that as a 'power failure' signal. This trick works for cheap, obscure or basic *dumb* UPSs.
jh _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Another solution could be cheap router with IP not pluged into UPS. If server can not ping that IP, you would shut it down, via script.
Ljubomir
Ljubomir Ljubojevic wrote:
Robert Heller wrote:
At Fri, 1 Jul 2011 12:26:10 +0100 (BST) CentOS mailing list centos@centos.org wrote:
On Fri, 1 Jul 2011, Timothy Murphy wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
<snip>
With a non-Linux compatable UPS, you can use a old analog serial modem
<snip>
Another solution could be cheap router with IP not pluged into UPS. If server can not ping that IP, you would shut it down, via script.
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
mark
On Fri, 1 Jul 2011, m.roth@5-cent.us wrote:
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
Only if it speaks the right language which doesn't seem to be guaranteed. apcupsd didn't have a clue about a Liebert UPS I tried it with.
jh
John Hodrien wrote:
On Fri, 1 Jul 2011, m.roth@5-cent.us wrote:
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
Only if it speaks the right language which doesn't seem to be guaranteed. apcupsd didn't have a clue about a Liebert UPS I tried it with.
Have you reported this to apcupsd developers?
Ljubomir
On Fri, 1 Jul 2011, Ljubomir Ljubojevic wrote:
John Hodrien wrote:
On Fri, 1 Jul 2011, m.roth@5-cent.us wrote:
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
Only if it speaks the right language which doesn't seem to be guaranteed. apcupsd didn't have a clue about a Liebert UPS I tried it with.
Have you reported this to apcupsd developers?
This was some years ago, but no, I don't think I did.
jh
Ljubomir Ljubojevic wrote:
John Hodrien wrote:
On Fri, 1 Jul 2011, m.roth@5-cent.us wrote:
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
Only if it speaks the right language which doesn't seem to be guaranteed.apcupsd didn't have a clue about a Liebert UPS I tried it with.
Have you reported this to apcupsd developers?
They may not be that interested; I mean, it *is* APC UPS daemon....
John Hodrien wrote:
On Fri, 1 Jul 2011, m.roth@5-cent.us wrote:
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
Only if it speaks the right language which doesn't seem to be guaranteed. apcupsd didn't have a clue about a Liebert UPS I tried it with.
Ok, that's an interesting datapoint that I need to file away, for the next UPS I buy. Thanks for the info.
mark
At Fri, 1 Jul 2011 09:23:31 -0400 CentOS mailing list centos@centos.org wrote:
Ljubomir Ljubojevic wrote:
Robert Heller wrote:
At Fri, 1 Jul 2011 12:26:10 +0100 (BST) CentOS mailing list centos@centos.org wrote:
On Fri, 1 Jul 2011, Timothy Murphy wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
<snip> >> With a non-Linux compatable UPS, you can use a old analog serial modem <snip> > Another solution could be cheap router with IP not pluged into UPS. If > server can not ping that IP, you would shut it down, via script.
*shrug* I think all the UPSs I've seen for consumers in the last five years seem to have a USB port to go to the computer. That, and apcupsd, are all you need.
APC UPSes are supported by apcupsd. Other brands, not so much. Some (read: cheaper models) have their own special protocol and don't include Linux support. These solutions are intended for the cheaper or otherwise 'unsupported' UPSes. It *sounds* like the OP does not need something smart and is probably looking for something cheap.
mark
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 7/1/2011 10:59 AM, Robert Heller wrote:
APC UPSes are supported by apcupsd. Other brands, not so much. Some (read: cheaper models) have their own special protocol and don't include Linux support. These solutions are intended for the cheaper or otherwise 'unsupported' UPSes. It *sounds* like the OP does not need something smart and is probably looking for something cheap.
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were 5-10 years ago. $250-$300 to protect $2000-$6000 worth of hardware is worth it in my book.
(I prefer the Smart-UPS units for a variety of reasons. Line filtering, voltage regulation, and nice reporting features via apcupsd. We have MRTG polling the apcupsd data regularly and have graphs of line voltage / operating temperature. There are even variants with the audible alarm disabled, which is perfect for a home office where you don't need that high powered screech.)
centos-bounces@centos.org wrote:
On 7/1/2011 10:59 AM, Robert Heller wrote:
APC UPSes are supported by apcupsd. Other brands, not so much. Some (read: cheaper models) have their own special protocol and don't include Linux support. These solutions are intended for the cheaper or otherwise 'unsupported' UPSes. It *sounds* like the OP does not need something smart and is probably looking for something cheap.
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were
5-10
years ago. $250-$300 to protect $2000-$6000 worth of hardware is
worth
it in my book.
To what extent does a UPS *protect* the hardware? Maintaining up-time during brief brown-outs is one thing I expect of a UPS, Orderly shutdown is another thing I expect of a UPS.
*protection* of the PC from irregularity in the AC Mains by a UPS, however, I question. Rather, it seems, any power irregularity that would kill a PC by propagating through the PSU will also propagate through the UPS.
NO UPS MADE TODAY (according to my reading of the stats on advertisements) eats lightning strikes and asks for more.
So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Insert spiffy .sig here: Life is complex: it has both real and imaginary parts. Life is not measured by the number of breaths we take, but by the moments that take our breath away.
//me ******************************************************************* This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. This footnote also confirms that this email message has been swept for the presence of computer viruses. www.Hubbell.com - Hubbell Incorporated**
Brunner, Brian T. wrote:
centos-bounces@centos.org wrote:
On 7/1/2011 10:59 AM, Robert Heller wrote:
APC UPSes are supported by apcupsd. Other brands, not so much. Some (read: cheaper models) have their own special protocol and don't include Linux support. These solutions are intended for the cheaper or otherwise 'unsupported' UPSes. It *sounds* like the OP does not need something smart and is probably looking for something cheap.
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were 5-10 years ago. $250-$300 to protect $2000-$6000 worth of hardware is
worth it in my book.
To what extent does a UPS *protect* the hardware? Maintaining up-time during brief brown-outs is one thing I expect of a UPS,Orderly shutdown is another thing I expect of a UPS.
*protection* of the PC from irregularity in the AC Mains by a UPS, however, I question. Rather, it seems, any power irregularity that would kill a PC by propagating through the PSU will also propagate through the UPS.
NO UPS MADE TODAY (according to my reading of the stats on advertisements) eats lightning strikes and asks for more.
So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Really? That's what you read in the specs? Here, I thought that good quality surge protectors would do that, and my UPS does says surge protection as well as UPS. IIRC, UPSs, and better surge protectors, offer a multi-thousand dollar warranty if it doesn't stop a large surge and your system's fried.
I *think* I have one of these, http://www.cyberpowersystems.com/products/ups-systems/soho-ups.html, which they say is suitable for SOHO usage... and you notice the "connected equipment guarantee" (CEG): between $25k USDand $100k USD, depending on model.
mark
centos-bounces@centos.org wrote:
Brunner, Brian T. wrote:
centos-bounces@centos.org wrote:
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were 5-10 years ago. $250-$300 to protect $2000-$6000 worth
of hardware is worth it in my book.
To what extent does a UPS *protect* the hardware? Maintaining up-time during brief brown-outs is one thing I expect of a UPS,Orderly shutdown is another thing I expect of a UPS.
*protection* of the PC from irregularity in the AC Mains by a UPS, however, I question. Rather, it seems, any power irregularity that would kill a PC by propagating through the PSU will also propagate through the UPS.
NO UPS MADE TODAY (according to my reading of the stats on advertisements) eats lightning strikes and asks for more.
So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Really? That's what you read in the specs?
Yes. Compare the joules rating (as being stopped by a UPS with surge suppression) to the joules required to damage the computer on the other side of a PSU. So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Robert Heller suggested that UPS architecture matters: AC->DC::DC Batteries::DC->AC Where input AC is electrically decoupled from output AC. Not many adverts for UPS's explain whether this is the case with their UPS.
Here, I thought that good quality surge protectors would do that, and my UPS does says surge protection as well as UPS. IIRC, UPSs, and
better surge
protectors, offer a multi-thousand dollar warranty if it doesn't stop
a large
surge and your system's fried.
Have you tried collecting on said warranties? Be prepared to prove that the surge that fried your hardware did not exceed the joules rating on the surge suppressor.
Insert spiffy .sig here: Life is complex: it has both real and imaginary parts. Life is not measured by the number of breaths we take, but by the moments that take our breath away.
//me ******************************************************************* This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. This footnote also confirms that this email message has been swept for the presence of computer viruses. www.Hubbell.com - Hubbell Incorporated**
On 7/1/2011 1:02 PM, Brunner, Brian T. wrote:
centos-bounces@centos.org wrote:
Brunner, Brian T. wrote:
centos-bounces@centos.org wrote:
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were 5-10 years ago. $250-$300 to protect $2000-$6000 worth
of hardware is worth it in my book.
To what extent does a UPS *protect* the hardware? Maintaining up-time during brief brown-outs is one thing I expect of a UPS,Orderly shutdown is another thing I expect of a UPS.
*protection* of the PC from irregularity in the AC Mains by a UPS, however, I question. Rather, it seems, any power irregularity that would kill a PC by propagating through the PSU will also propagate through the UPS.
NO UPS MADE TODAY (according to my reading of the stats on advertisements) eats lightning strikes and asks for more.
So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Really? That's what you read in the specs?
Yes. Compare the joules rating (as being stopped by a UPS with surge suppression) to the joules required to damage the computer on the other side of a PSU. So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Robert Heller suggested that UPS architecture matters: AC->DC::DC Batteries::DC->AC Where input AC is electrically decoupled from output AC. Not many adverts for UPS's explain whether this is the case with their UPS.
APC's SmartUPS line, Liebert, and Eaton Powerware are all true-sine wave UPS's, and do proper decoupling. Unfortunately, this kind of data doesn't make for great ad copy, so it's left out, and you have to dig deep into datasheets to get that information. I pretty much only use APC, and we have truly crap power here. Because of some heavy industry in the area, brownouts are common, and that'll kill a PC power supply better than anything. I've pulled one 7 year old APC from a server closet where the lightning took the top of the telephone pole OFF. THE UPS was fried, some of the breakers in the building were fused (!), but the servers were fine, outside of the router that got zapped from the DSL modem.
The advantage to better UPS systems is they dump the input power through a big, beefy transformer. That provides enough of an inductor that it can eat a HUGE surge before the insulation in the transformer breaks down and it arcs across to the output. Even then, it has a long way to go before it can hit the output circuits.
So, your cheap $100 UPS won't provide as much protection from a nasty spike, but it would be VERY rare to see a spike that big.
Here, I thought that good quality surge protectors would do that, and my UPS does says surge protection as well as UPS. IIRC, UPSs, and
better surge
protectors, offer a multi-thousand dollar warranty if it doesn't stop
a large
surge and your system's fried.
Have you tried collecting on said warranties? Be prepared to prove that the surge that fried your hardware did not exceed the joules rating on the surge suppressor.
And that's why you don't go cheap on your UPSs. Overbuilding capacity means you get a longer run time - and you also have room for expansion. Cheaper to go big early.
Insert spiffy .sig here: Life is complex: it has both real and imaginary parts. Life is not measured by the number of breaths we take, but by the moments that take our breath away.
//me
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. This footnote also confirms that this email message has been swept for the presence of computer viruses. www.Hubbell.com - Hubbell Incorporated**
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Robert Heller suggested that UPS architecture matters: AC->DC::DC Batteries::DC->AC Where input AC is electrically decoupled from output AC. Not many adverts for UPS's explain whether this is the case with their UPS.
APC's SmartUPS line, Liebert, and Eaton Powerware are all true-sine wave UPS's, and do proper decoupling. Unfortunately, this kind of data doesn't make for great ad copy, so it's left out, and you have to dig deep into datasheets to get that information. I pretty much only use APC, and we have truly crap power here. Because of some heavy industry in the area, brownouts are common, and that'll kill a PC power supply better than anything. I've pulled one 7 year old APC from a server closet where the lightning took the top of the telephone pole OFF. THE UPS was fried, some of the breakers in the building were fused (!), but the servers were fine, outside of the router that got zapped from the DSL modem.
I beg to differ about APC. The accepted term for what Robert described is a "double conversion" or "online" UPS. APC's SmartUPS family is only available with the double conversion feature if you specify a "SmartUPS Online" model. The rest of the SmartUPS family use "Line Interactive" which runs on mains power until the voltage/current/frequency goes out of tolerance, at which point they cut over to battery. The Liebert GXT2/3 family which we use quite a bit of were, until recently, strictly double conversion.
At Fri, 1 Jul 2011 11:46:40 -0400 CentOS mailing list centos@centos.org wrote:
centos-bounces@centos.org wrote:
On 7/1/2011 10:59 AM, Robert Heller wrote:
APC UPSes are supported by apcupsd. Other brands, not so much. Some (read: cheaper models) have their own special protocol and don't include Linux support. These solutions are intended for the cheaper or otherwise 'unsupported' UPSes. It *sounds* like the OP does not need something smart and is probably looking for something cheap.
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were
5-10
years ago. $250-$300 to protect $2000-$6000 worth of hardware is
worth
it in my book.
To what extent does a UPS *protect* the hardware? Maintaining up-time during brief brown-outs is one thing I expect of a UPS, Orderly shutdown is another thing I expect of a UPS.
*protection* of the PC from irregularity in the AC Mains by a UPS, however, I question. Rather, it seems, any power irregularity that would kill a PC by propagating through the PSU will also propagate through the UPS.
A *good* UPS has a surge protector, then a good filtering power supply, which functions as a battery charger. Then there is an inverter (powered by the battery) that generates 'fresh' AC. 'Normal' surges are soaked up by the input 'battery charger' supply. A UPS that decouples the line power from its output by using the inverter all of the (not just during a power failure) effectively isolates the output from the the input -- all irregularities in the AC Mains are absorbed by the battery charger supply. The battery is very tolerant and does not need the battery charger circuit to provide a *precise* continious voltage -- eg dropouts that take the battery charger circuit 'off line' for miliseconds are not going to affect the battery. Nor will modest surges (regulation in the battery charger circuit should take care of larger surges and MOVs on the AC Mains should take care of really large surges). The inverter will be powered by the battery charger or battery, depending on which is functioning at any given instant, and the input to the inverter will be a flat, smooth DC voltage in either case.
NO UPS MADE TODAY (according to my reading of the stats on advertisements) eats lightning strikes and asks for more.
It is likely that the UPS would die, leaving the computer, etc. untouched.
So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Certainly random 'low-level' surges (typical 'dirty' power as provided by the power company). The UPS would also be the front-line 'cannon fodder' for more massive surges (eg lightning strikes).
In reality, a *properly* wired building (one that is up to code), will have effective lightning protection as part of the basic wiring. A data center wiring will be even better.
Insert spiffy .sig here: Life is complex: it has both real and imaginary parts. Life is not measured by the number of breaths we take, but by the moments that take our breath away.
//me
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. This footnote also confirms that this email message has been swept for the presence of computer viruses. www.Hubbell.com - Hubbell Incorporated**
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, Jul 1, 2011 at 5:46 PM, Brunner, Brian T. BBrunner@gai-tronics.com wrote:
NO UPS MADE TODAY (according to my reading of the stats on advertisements) eats lightning strikes and asks for more.
So per your experiences and greater technical savvy: What PSU/PC kill power irregularities will be stopped by which UPS?
Well, the UPS itself normally blows a fuse, or itself gets burnt out, so you pay a small price to protect a bigger investment
On Friday, July 01, 2011 11:46 PM, Brunner, Brian T. wrote:
centos-bounces@centos.org wrote:
On 7/1/2011 10:59 AM, Robert Heller wrote:
APC UPSes are supported by apcupsd. Other brands, not so much. Some (read: cheaper models) have their own special protocol and don't include Linux support. These solutions are intended for the cheaper or otherwise 'unsupported' UPSes. It *sounds* like the OP does not need something smart and is probably looking for something cheap.
And the APC Smart-UPS 750 units are not all that expensive either. Even the 1500VA units are a lot less expensive then they were
5-10
years ago. $250-$300 to protect $2000-$6000 worth of hardware is
worth
it in my book.
To what extent does a UPS *protect* the hardware? Maintaining up-time during brief brown-outs is one thing I expect of a UPS, Orderly shutdown is another thing I expect of a UPS.
*protection* of the PC from irregularity in the AC Mains by a UPS, however, I question. Rather, it seems, any power irregularity that would kill a PC by propagating through the PSU will also propagate through the UPS.
PSUs must love Regular under voltage electricity and so too your data if the batteries of the UPS are anything to go by. Batteries died within two years on one particular circuit and the connected servers suffered while I was getting replacement batteries. Apparently one motherboard loved it so much that the thing would not POST anymore.
On Fri, Jul 1, 2011 at 1:05 PM, Timothy Murphy gayleard@eircom.net wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
My server, an HP MicroServer, came back (re-booted) on 2 of the 3 occasions, but not on the third.
I assume that the problem arises because the machine does not close down properly. (Although it is also possible that a voltage surge might have been responsible - I have no surge protector on this supply.)
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
I know there is a Remote Management (iLO) card for this machine, which might be useful for this. Unfortunately, I've already used the PCIe slot for a second ethernet card.
Any advice or suggestions gratefully received.
-- Timothy Murphy e-mail: gayleard /at/ eircom.net tel: +353-86-2336090, +353-1-2842366 s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
A UPS would be your simplest option here since the UPS can send a signal to the OS to shutdown properly.
Using a "torch battery" (I presume this is a large torch?) you'll still have the same issue as you have now - when the battery runs flat (i.e. power outage is longer than 10 minutes or so) Linux will still crash uncleanly.
Rudi Ahlers wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
A UPS would be your simplest option here since the UPS can send a signal to the OS to shutdown properly.
Using a "torch battery" (I presume this is a large torch?) you'll still have the same issue as you have now - when the battery runs flat (i.e. power outage is longer than 10 minutes or so) Linux will still crash uncleanly.
As will be obvious, I know nothing in this area. My thought was just that the machine only requires say 30 seconds of life to shutdown properly, and I would have thought there was enough capacity in a large torch battery to supply this?
But there seems to be a 100% backing for UPS, so I'll look into that.
On 7/1/2011 12:36 PM, Timothy Murphy wrote:
Rudi Ahlers wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
A UPS would be your simplest option here since the UPS can send a signal to the OS to shutdown properly.
Using a "torch battery" (I presume this is a large torch?) you'll still have the same issue as you have now - when the battery runs flat (i.e. power outage is longer than 10 minutes or so) Linux will still crash uncleanly.
As will be obvious, I know nothing in this area. My thought was just that the machine only requires say 30 seconds of life to shutdown properly, and I would have thought there was enough capacity in a large torch battery to supply this?
But there seems to be a 100% backing for UPS, so I'll look into that.
A UPS is the implementation of your "torch battery" idea. Remember that the battery will need a way to convert from DC to the A/C power required by your machine. There also needs to be circuitry to switch over to battery power in the event of an outage. You also need to make sure the battery stays charged. The UPS does all of this for you. The UPS can also notify your server in the event of a power outage so it can shut down cleanly. If you don't require a long runtime, then you don't need to get a huge UPS. APC's website has a calculator that can help you determine which UPS will work best based on your equipment and desired runtime.
Bowie Bailey Bowie_Bailey@BUC.com wrote:
If you don't require a long runtime, then you don't need to get a huge UPS. APC's website has a calculator that can help you determine which UPS will work best based on your equipment and desired runtime.
However, give yourself some leeway to allow not only for changes in your usage pattern (addition of hardware or changing hardware) but also degradation of the battery over time.
I've generally found that if you're pushing a UPS past 70-85% capacity, you may not (depending on the system) have sufficient runtime to shut down completely before exhausting the battery. Other things to take into account is if you have programs that take a while to shut down (some databases can take a *very* long time) and whether or not you have to power auxilliary equipment such as monitor during an outage.
I usually like to have a UPS loaded to only 50%. That gives a decent enough battery time for either allowing an extended shutdown time or for carrying the system through multiple short failures without bringing the system down at all.
There is one flaw that I know of with APC brand UPSes, although I wouldn't be surprised if other UPSes are similar (since APC has traditionally set the standard in the market): There is a small window between the time that the UPS initiates a shutdown and the control software (such as apcupsd) tells the UPS to kill power in (some number of) seconds. If your mains power returns and stays on during that window, your machine may not restart by itself: It may be too late to cancel the OS shutdown and subsequent power-off, but because the UPS still has mains power it doesn't trigger the power cycle that allows the computer to reboot (assuming you've configured your BIOS to boot at power-on). In my experience this doesn't happen often (I've seen it probably twice in 15 years), but it does happen and can be a real PITA if you don't have anyone on site that can deal with it.
Devin
On 07/02/2011 12:07 PM, Devin Reade wrote:
There is one flaw that I know of with APC brand UPSes, although I wouldn't be surprised if other UPSes are similar (since APC has traditionally set the standard in the market): There is a small window between the time that the UPS initiates a shutdown and the control software (such as apcupsd) tells the UPS to kill power in (some number of) seconds. If your mains power returns and stays on during that window, your machine may not restart by itself: It may be too late to cancel the OS shutdown and subsequent power-off, but because the UPS still has mains power it doesn't trigger the power cycle that allows the computer to reboot (assuming you've configured your BIOS to boot at power-on). In my experience this doesn't happen often (I've seen it probably twice in 15 years), but it does happen and can be a real PITA if you don't have anyone on site that can deal with it.
Fix the shutdown sequence so that after killing the UPS, and right at the point where you would tell the BIOS to turn off the ATX power supply, the machine sleeps for a few seconds (longer than it should take for power to drop) and then reboots instead of shutting down. If commercial power has returned, you just reboot. If not, power is removed at a safe point.
Devin Reade wrote:
There is one flaw that I know of with APC brand UPSes, although I wouldn't be surprised if other UPSes are similar (since APC has traditionally set the standard in the market): There is a small window between the time that the UPS initiates a shutdown and the control software (such as apcupsd) tells the UPS to kill power in (some number of) seconds. If your mains power returns and stays on during that window, your machine may not restart by itself: It may be too late to cancel the OS shutdown and subsequent power-off, but because the UPS still has mains power it doesn't trigger the power cycle that allows the computer to reboot (assuming you've configured your BIOS to boot at power-on). In my experience this doesn't happen often (I've seen it probably twice in 15 years), but it does happen and can be a real PITA if you don't have anyone on site that can deal with it.
This can be solved with several approaches.
First is to have LAN power controller/switch, can't remember how you call it. It can monitor several inputs like voltage, on/off, etc, and it can be used to cut and restore power, as well as reboot. They have web interface and are accessible via IP. Wireless network guys use them.
Another approach is to have power switch power by mobile/cell phone. You call the number and it effectively cuts power for a desired period of time.
Ljubomir
Ljubomir Ljubojevic office@plnet.rs wrote:
First is to have LAN power controller/switch, can't remember how you call it. It can monitor several inputs like voltage, on/off, etc, and it can be used to cut and restore power, as well as reboot. They have web interface and are accessible via IP. Wireless network guys use them.
I do in fact use these (they're called managed power distribution units, such as the APC AP7900), not only for such manual operations but also because they're good for doing fencing in HA clusters.
However, that does not alter the fact that this solution requires human intervention (albiet in this case you can do it remotely). It doesn't fix the flaw, it just works around it.
It's also an additional expense that I don't think would be in the OP's scope.
Devin
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Devin Reade Sent: Saturday, July 02, 2011 15:51 To: CentOS mailing list Subject: Re: [CentOS] Power-outage
Ljubomir Ljubojevic office@plnet.rs wrote:
First is to have LAN power controller/switch, can't
remember how you
call it. It can monitor several inputs like voltage,
on/off, etc, and
it can be used to cut and restore power, as well as reboot.
They have
web interface and are accessible via IP. Wireless network
guys use them.
I do in fact use these (they're called managed power distribution units, such as the APC AP7900), not only for such manual operations but also because they're good for doing fencing in HA clusters.
However, that does not alter the fact that this solution requires human intervention (albiet in this case you can do it remotely). It doesn't fix the flaw, it just works around it.
Our dhcp server can run for many days on battery, it has a script running that if a machine does not ping, it sends a WOL. It does not do this if the network AC is off. If the AC is resored for more than 30 minutes, the script resumes. Machines wake up.
It's also an additional expense that I don't think would be in the OP's scope.
Also regarding DB taking too long to shut down, we use hibernation now. It is fast.
-Jason
-- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - - - Jason Pyeron PD Inc. http://www.pdinc.us - - Principal Consultant 10 West 24th Street #100 - - +1 (443) 269-1555 x333 Baltimore, Maryland 21218 - - - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- This message is copyright PD Inc, subject to license 20080407P00.
At Fri, 01 Jul 2011 17:36:56 +0100 CentOS mailing list centos@centos.org wrote:
Rudi Ahlers wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
A UPS would be your simplest option here since the UPS can send a signal to the OS to shutdown properly.
Using a "torch battery" (I presume this is a large torch?) you'll still have the same issue as you have now - when the battery runs flat (i.e. power outage is longer than 10 minutes or so) Linux will still crash uncleanly.
As will be obvious, I know nothing in this area. My thought was just that the machine only requires say 30 seconds of life to shutdown properly, and I would have thought there was enough capacity in a large torch battery to supply this?
What do you mean by 'torch battery'? If you mean the D cells typical of a flashlight (flashlight is 'American' English for [electric] torch), no this is not going to power a computer for more than fraction of a second. If you mean a 'lattern battery' (a larger 6V battery, used in typical upright camping latterns), maybe, but you are going to need an inverter, etc. to feed AC to your computer's power supply.
But there seems to be a 100% backing for UPS, so I'll look into that.
Yes. They are not terribly expensive, partitularly if you are not looking for a large one or a 'fancy' (excessive 'bells and whistles') one. For a small server, a typical consumer-grade UPS costing like US$100 will do everything you need.
These little consumer-grade UPSes, basically consist of a rechargable 'lattern battery' sized rechargable battery (commonly a gel-cell lead-acid type battery), a charger for the battery, and an inverter to re-create the AC power for your equipment (computer). All in one box. Most now have some simple 'smart' electronics with a simple micro-processor element with a USB connection that will talk to your computer telling it how things are going (on Mains, on battery, battery charging, battery charged, battery discharged, current load levels, Mains voltage, etc.).
On Fri, Jul 1, 2011 at 7:05 AM, Timothy Murphy gayleard@eircom.net wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
My server, an HP MicroServer, came back (re-booted) on 2 of the 3 occasions, but not on the third.
I assume that the problem arises because the machine does not close down properly. (Although it is also possible that a voltage surge might have been responsible - I have no surge protector on this supply.)
I've seen this happen before. The machine looses power long enough for the system to hang as the proper voltage is not maintained, but not long enough for it to turn off. A cheap UPS is what you need. Just something to smooth out the momentary power faults so the machine can shutdown or restart. A APC Back-UPs would be perfect and shouldn't break the bank. You don't need an expensive sinewave output like the APC Smart-UPS for what you are trying to accomplish.
Ryan
Ryan Wagoner wrote:
On Fri, Jul 1, 2011 at 7:05 AM, Timothy Murphy gayleard@eircom.net wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
<snip>
(Although it is also possible that a voltage surge might have been responsible - I have no surge protector on this supply.)
Speaking as someone who's lived in Chicago, central Florida, and the Washington, DC metro area, DON'T trust the electricity, always have your machine on a cheap surge protector, at least. <snip>
long enough for it to turn off. A cheap UPS is what you need. Just something to smooth out the momentary power faults so the machine can shutdown or restart. A APC Back-UPs would be perfect and shouldn't
<snip> Any of them. I've got a CyberPower at home, and have had another brand, and they *all* have a USB connection; none cost me more, over the last 10 years, than about $70 US.
Nothing against APC - we have a ton of them (literally, or more) at work - it's just that these were cheaper by 10%-25%.
mark
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
Colin.
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
I agree. Esp. since, other than in datacenters, *most* electric power is pretty crappy.
mark "let's not discuss ComEd in Chicago"
-------- Original Message -------- Subject: Re: [CentOS] Power-outage From: m.roth@5-cent.us To: CentOS mailing list centos@centos.org Date: Friday, July 01, 2011 9:28:21 AM
Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
I agree. Esp. since, other than in datacenters, *most* electric power is pretty crappy.
mark "let's not discuss ComEd in Chicago"
I would have to disagree. They probably put high efficiency active PFC power supplies in the servers to save YOU money. You could buy a cheaper PSU that will not be as efficient and would thus cost you more in electric costs and create more heat (which would again cost you more in AC bills and reduce server density). The active PFC supplies are actually better at dealing with high/low voltages, however they do require actual AC power that conforms to a true sine wave. Newer/better UPS units output sine waves, cheaper or older UPS units may only output approximated (aka stepped) sine waves.
Dell has done this in some of their boxes too, and I would expect to see it occur more often as more consumers are looking at 80+ and better certified PSUs.
Blake Hudson wrote:
From: m.roth@5-cent.us
Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
I agree. Esp. since, other than in datacenters, *most* electric power is pretty crappy.
I would have to disagree. They probably put high efficiency active PFC power supplies in the servers to save YOU money. You could buy a cheaper PSU that will not be as efficient and would thus cost you more in electric costs and create more heat (which would again cost you more in AC bills and reduce server density). The active PFC supplies are
Except that I expect datacenters to have conditioned power, and so they can cheap out with the servers, with the same expectations. And I would expect consumer-grade systems to not have fancy power units, but ones that won't die on power irregularities from the electric co's. <snip> mark
-------- Original Message -------- Subject: Re: [CentOS] Power-outage From: m.roth@5-cent.us To: CentOS mailing list centos@centos.org Date: Friday, July 01, 2011 9:57:41 AM
Blake Hudson wrote:
From: m.roth@5-cent.us
Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
I agree. Esp. since, other than in datacenters, *most* electric power is pretty crappy.
I would have to disagree. They probably put high efficiency active PFC power supplies in the servers to save YOU money. You could buy a cheaper PSU that will not be as efficient and would thus cost you more in electric costs and create more heat (which would again cost you more in AC bills and reduce server density). The active PFC supplies are
Except that I expect datacenters to have conditioned power, and so they can cheap out with the servers, with the same expectations. And I would expect consumer-grade systems to not have fancy power units, but ones that won't die on power irregularities from the electric co's.
<snip> mark
I think you missed the point - While manufacturer's could (and probably sometimes do) "cheap out with the servers" power supplies, it is not in your best interest (or their's).
More efficient PSUs create less waste heat and draw less power which means higher density, more performance, etc. This is more important in the server space where the computers are on 24/7 and tightly packed into racks. More efficient PSUs cost more upfront than inefficient ones, which mean that Dell/HP/etc can probably make a higher profit. In the long term, you may be saving $50-100 per server per year on reduced electric and associated costs. If you're in a colo with power draw restrictions, you may be saving even more.
Blake Hudson wrote:
From: m.roth@5-cent.us
Blake Hudson wrote:
From: m.roth@5-cent.us
Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote: > Any advice or suggestions gratefully received. If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
I agree. Esp. since, other than in datacenters, *most* electric power is pretty crappy.
I would have to disagree. They probably put high efficiency active PFC power supplies in the servers to save YOU money. You could buy a cheaper PSU that will not be as efficient and would thus cost you more in electric costs and create more heat (which would again cost you more in AC bills and reduce server density). The active PFC supplies are
Except that I expect datacenters to have conditioned power, and so they can cheap out with the servers, with the same expectations. And I would expect consumer-grade systems to not have fancy power units, but ones that won't die on power irregularities from the electric co's.
<snip>
I think you missed the point - While manufacturer's could (and probably sometimes do) "cheap out with the servers" power supplies, it is not in your best interest (or their's).
Oh, I haven't missed the point - you missed my point, that they will cheap out - that's in the interest of their stockholders, and their exec's stock options. Better power supplies, though they actually cost *them* a few dollars more, are much more expensive options.
More efficient PSUs create less waste heat and draw less power which means higher density, more performance, etc. This is more important in the server space where the computers are on 24/7 and tightly packed into racks. More efficient PSUs cost more upfront than inefficient ones, which mean that Dell/HP/etc can probably make a higher profit. In the long term, you may be saving $50-100 per server per year on reduced electric and associated costs. If you're in a colo with power draw restrictions, you may be saving even more.
Yup. As a matter of fact, my own brand new machine here at work just arrived an hour ago (ah, the smell of fresh plastic outgassing, factory air from China :((( ), and when I spec'd it out, this was the option:
Precision T3500 CMT Standard PSU, C2 Motherboard [Included in Price] Precision T3500, CMT, 85 Percent Efficient Power Supply, C2 Motherboard [add $42.82]
You know that's maybe $5, with the quantities they're buying.
mark
On Friday 01 July 2011 15:25, Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
Lock-in tactics I think, I ended up having to buy HP USP's at about 5x the price.
Colin
Colin Coles wrote:
On Friday 01 July 2011 15:25, Nicolas Thierry-Mieg wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
Lock-in tactics I think, I ended up having to buy HP USP's at about 5x the price.
Question: are we talking server-grade systems, rackmounts? I can't imagine that they'd do that for consumer-grade machines.
mark
At Fri, 01 Jul 2011 16:25:33 +0200 CentOS mailing list centos@centos.org wrote:
Colin Coles wrote:
On Friday 01 July 2011 12:05, Timothy Murphy wrote:
Any advice or suggestions gratefully received.
If you are thinking of the UPS route a caveat: I have several HP servers and most of them will not work on cheap UPS's as they do not produce the pure sine wave modern HP machines require but rather a crude stepped voltage.
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
AND *I* thought *switching power supplies* (effectively) rectified the AC input and then used the DC to drive a higher frequency system to get the desired output voltages. (The higher frequency means smaller, more efficient transformers and need smaller filter caps -- all of which means a lower cost, more reliable, more efficient power supply.) Which suggests that both the input voltage and frequency are not particularly critical, so long as it does not have massive spikes/surges or consistently low voltage.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 7/1/11, Robert Heller heller@deepsoft.com wrote:
At Fri, 01 Jul 2011 16:25:33 +0200 CentOS mailing list centos@centos.org wrote:
Colin Coles wrote:
perhaps naively, I'm surprised: doesn't this mean they put crappy PSUs in those servers? I thought decent PSUs were expected to deal with dirty input AC?
AND *I* thought *switching power supplies* (effectively) rectified the AC input and then used the DC to drive a higher frequency system to get the desired output voltages. (The higher frequency means smaller, more efficient transformers and need smaller filter caps -- all of which means a lower cost, more reliable, more efficient power supply.) Which suggests that both the input voltage and frequency are not particularly critical, so long as it does not have massive spikes/surges or consistently low voltage.
In a normal SMPS that would be true, because the typical ATX PSU normally has two bulk input capacitors. However the better PSUs nowadays and those in servers are usually active PFC units which only has one.
Now the problem occurs because non-true sinewave UPS usually use PWM to achieve the output. This means the UPS outputs a consistent high voltage but switches it on/off to achieve the same average power, i.e. 400V for x msec, then 0V for x msec = 200V average, where x should be much smaller than 16 msec IIRC, which is the required hold up time for ATX specifications.
In a cheap and arguably badly designed UPS, the selected voltage is much higher than the PSU is expected to ever handle from a true sine-wave source (which is nominally 320V peak for a 230V RMS source). So it either blows the input capacitors (typically 400V values), or protective circuitry shuts it down first. On the non-PFC PSU, the voltage is divided across the two main caps so this isn't a problem.
On 7/1/11, Timothy Murphy gayleard@eircom.net wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
Like others have suggested, a cheap UPS is the way to go. The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and something to hold the batteries (you'll need more than one because attempting to draw a huge current from a normal battery will either kill it or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
Like others have suggested, a cheap UPS is the way to go.
I'm convinced. Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and something to hold the batteries (you'll need more than one because attempting to draw a huge current from a normal battery will either kill it or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
Many decades ago I went to lectures at university given by Fred Hoyle (famous at the time for a TV series where he said God was unnecessary). The lectures (on thermodynamics) were not really very good, but they were interesting because Fred Hoyle was slighly paranoid, and believed evil capitalists were foisting unnecessary devices on us.
One of his pet theories was that cars did not need huge accumulators, but could be started with a torch-battery.
Another was that incandescent bulbs were deliberately made to fail after a certain time.
Another was razor blades, which according to him could easily last for ever.
One interesting idea was that instead of nuclear power stations it would be cheaper, and give the same energy, to plant trees in a strip around the equator (I forget how wide).
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Timothy Murphy Sent: Saturday, July 02, 2011 8:52 To: centos@centos.org Subject: Re: [CentOS] Power-outage
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine
alive long
enough to make a graceful exit.
Like others have suggested, a cheap UPS is the way to go.
I'm convinced. Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
http://www.amazon.com/APC-Back-UPS-shutdown-software-UPS-BE350G/dp/B001985SW...
The problem with your idea is that you'll need a DC to AC inverter that
can handle
the output current required by your server and something to
hold the
batteries (you'll need more than one because attempting to
draw a huge
current from a normal battery will either kill it or at the
very least
cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money
and time to
have this thing.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
Many decades ago I went to lectures at university given by Fred Hoyle (famous at the time for a TV series where he said God was unnecessary). The lectures (on thermodynamics) were not really very good, but they were interesting because Fred Hoyle was slighly paranoid, and believed evil capitalists were foisting unnecessary devices on us.
One of his pet theories was that cars did not need huge accumulators, but could be started with a torch-battery.
Another was that incandescent bulbs were deliberately made to fail after a certain time.
Another was razor blades, which according to him could easily last for ever.
One interesting idea was that instead of nuclear power stations it would be cheaper, and give the same energy, to plant trees in a strip around the equator (I forget how wide).
-- Timothy Murphy e-mail: gayleard /at/ eircom.net tel: +353-86-2336090, +353-1-2842366 s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Jason Pyeron wrote:
Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
BE350G/dp/B001985SWW/
Thanks, I'll look into that.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
Surely one 12v battery would do? It is only meant to last for 30 seconds or so, so wouldn't reducing the voltage be easy enough? I repeat that I don't know what I'm talking about ...
On Saturday, July 02, 2011 09:42 PM, Timothy Murphy wrote:
Jason Pyeron wrote:
Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
BE350G/dp/B001985SWW/
Thanks, I'll look into that.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
Surely one 12v battery would do? It is only meant to last for 30 seconds or so, so wouldn't reducing the voltage be easy enough? I repeat that I don't know what I'm talking about ...
The PSU transforms incoming electricity to various voltages on multiple rails. You need more than just a 12V lead acid battery.
At Sat, 02 Jul 2011 15:42:38 +0200 CentOS mailing list centos@centos.org wrote:
Jason Pyeron wrote:
Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
BE350G/dp/B001985SWW/
Thanks, I'll look into that.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
Surely one 12v battery would do?
Actually not. You need both positive and negative voltages. You cannot just use a 12v battery with a voltage divider resistor network. And because the voltages needed have to be precise, you really need a properly regulated power supply. Like Jason said, it would be a 48VDC input unit and will cost you many times what a cheap UPS would cost.
It is only meant to last for 30 seconds or so, so wouldn't reducing the voltage be easy enough? I repeat that I don't know what I'm talking about ...
On Sat, Jul 02, 2011 at 10:45:13AM -0400, Robert Heller wrote:
At Sat, 02 Jul 2011 15:42:38 +0200 CentOS mailing list centos@centos.org wrote:
Jason Pyeron wrote:
Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
BE350G/dp/B001985SWW/
Thanks, I'll look into that.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
Surely one 12v battery would do?
Actually not. You need both positive and negative voltages. You cannot just use a 12v battery with a voltage divider resistor network. And because the voltages needed have to be precise, you really need a properly regulated power supply. Like Jason said, it would be a 48VDC input unit and will cost you many times what a cheap UPS would cost.
It is only meant to last for 30 seconds or so, so wouldn't reducing the voltage be easy enough? I repeat that I don't know what I'm talking about ...
I believe Google runs bazillions of servers with a PS that emits 12 v, a 12v battery on the downstream side of the PS, and a custom motherboard that requires only 12V input. If not exactly those specs, at least that sort of thing. But Google uses so many tens of thousands of 'em that it is economical for them to have them custom-made, whereas you want one, so it'd cost you a fortune. (there have been a few articles on various web sites over the last year or two showing Google's server internals.)
So, in theory you could do something akin to what you ask, but in practice it would be much cheaper for you to spend a few bucks for a low-spec UPS unit. assuming the server has no peripherals that require power (such as a monitor) and it is a low-power device like a micro-ATX board with an Atom processor, or someting of that ilk, a small-ish UPS would work fine. Such low-spec UPS units are inexpensive.
On Saturday, July 02, 2011 09:00:54 AM Jason Pyeron wrote:
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
A company called PowerStream produces DC input ATX supplies for 12V, 24V, and 48V input, all with up to 500W of power. The 12V input page is at http://www.powerstream.com/DC-PC-12V.htm
We have a number of their -48V input supplies in use. No, the 500W version in 12V input is not cheap.
Lamar Owen wrote:
On Saturday, July 02, 2011 09:00:54 AM Jason Pyeron wrote:
You will either need many different batteries for the different voltages (1.2, 3.3, 5, 12, -12, -5) or a DC ATX power supply (not cheap and not very powerful until the 48V input variety)
A company called PowerStream produces DC input ATX supplies for 12V, 24V, and 48V input, all with up to 500W of power. The 12V input page is at http://www.powerstream.com/DC-PC-12V.htm
We have a number of their -48V input supplies in use. No, the 500W version in 12V input is not cheap.
There are smaller and cheaper 12V solutions Like the picoPSU's: http://www.mini-box.com/picoPSU-160-XT
Ljubomir
On Wednesday, July 06, 2011 05:23:32 AM Ljubomir Ljubojevic wrote:
Lamar Owen wrote:
We have a number of their -48V input supplies in use. No, the 500W version in 12V input is not cheap.
There are smaller and cheaper 12V solutions Like the picoPSU's: http://www.mini-box.com/picoPSU-160-XT
Sure; we have a couple of small units like that for some solar-powered things we're doing here; however, the max I've seen for those plug-in type small ATX/ITX power supplies has been in the ~200W range (the specific one you linked to is only 160W), and my reply was specifically directed at the idea that lower than 48 VDC input was limited in power handling...... PowerStream has a 500W 12VDC input unit, which is quite a bit more power than I've seen in the mini-ITX plugin supply categories.
On 07/06/11 5:44 AM, Lamar Owen wrote:
Sure; we have a couple of small units like that for some solar-powered things we're doing here; however, the max I've seen for those plug-in type small ATX/ITX power supplies has been in the ~200W range (the specific one you linked to is only 160W), and my reply was specifically directed at the idea that lower than 48 VDC input was limited in power handling...... PowerStream has a 500W 12VDC input unit, which is quite a bit more power than I've seen in the mini-ITX plugin supply categories.
500 watts at 12VDC is 41 amps. that requires some hefty wiring, and if you have to run it any distances, either the wire is ridiculously heavy (and expensive) or you suffer from voltage drop under load.
500 watts at 120V is only 4 amps, and can easily be run 100s of feet through simple lamp cord sized wiring.
On Wed, Jul 6, 2011 at 6:24 PM, John R Pierce pierce@hogranch.com wrote:
On 07/06/11 5:44 AM, Lamar Owen wrote:
Sure; we have a couple of small units like that for some solar-powered things we're doing here; however, the max I've seen for those plug-in type small ATX/ITX power supplies has been in the ~200W range (the specific one you linked to is only 160W), and my reply was specifically directed at the idea that lower than 48 VDC input was limited in power handling...... PowerStream has a 500W 12VDC input unit, which is quite a bit more power than I've seen in the mini-ITX plugin supply categories.
500 watts at 12VDC is 41 amps. that requires some hefty wiring, and if you have to run it any distances, either the wire is ridiculously heavy (and expensive) or you suffer from voltage drop under load.
500 watts at 120V is only 4 amps, and can easily be run 100s of feet through simple lamp cord sized wiring.
-- john r pierce N 37, W 122 santa cruz ca mid-left coast
Which is why it's generally better to use 48V for these kinds of applications :)
On Wednesday, July 06, 2011 12:24:12 PM John R Pierce wrote:
500 watts at 12VDC is 41 amps. that requires some hefty wiring, and if you have to run it any distances, either the wire is ridiculously heavy (and expensive) or you suffer from voltage drop under load.
While not CentOS-specific, this *is* in my area of expertise. We have 540Ah of -48VDC driven by a pair of Lorain Flotrol 200A rectifiers for our telco equipment (including the Cisco 12008 and OSR7609 routers). Our solar sites are mostly 24VDC with, again, 540Ah minimum at each site, with a few 12VDC systems with 75 to 300Ah at each site. I've run enough 4/0 and larger flex cables, that's for sure..... for 41 amps, up to 25 feet or so, relatively small 8AWG is sufficient. That's smaller gauge than the 6AWG and 4AWG I ran for the 12008 and 7609, respectively, for -48VDC power. (I say relatively small; the largest conductor size we have here is 6kA rated busbar, so even 2AWG or 2/0 AWG is relatively small......:-) )
I've seen much larger, specifically in the Brookhaven 5ESS in Atlanta. I remember seeing one branch circuit idling at ~2.5kA. Hmmm, speaking of 5ESS, I wonder what the chance of a CentOS for a 3B15 or 3B20 would be? :-) (No, Russ, before you ask: I don't still have the 3B15's that used to be here.....)
For a reference on DC power design, useful if you need to support CentOS servers with DC supplies in a telco environment, please see "DC Power System Design for Telecommunications" by Whitham D. Reeve for the 'canonical' reference work. Everything you need, including current limit and overcurrent protection, low-voltage cutouts, distribution design, voltage drop and wire sizing calculations, and ampacity tables for DC (NEC includes AC ampacity tables, but not DC).
And I have a few CentOS boxes running on DC power. And, of course, having powertop running on CentOS, and having some low-power modes, helps tremendously.
On Wed, 6 Jul 2011, John R Pierce wrote:
To: centos@centos.org From: John R Pierce pierce@hogranch.com Subject: Re: [CentOS] Power-outage
On 07/06/11 5:44 AM, Lamar Owen wrote:
Sure; we have a couple of small units like that for some solar-powered things we're doing here; however, the max I've seen for those plug-in type small ATX/ITX power supplies has been in the ~200W range (the specific one you linked to is only 160W), and my reply was specifically directed at the idea that lower than 48 VDC input was limited in power handling...... PowerStream has a 500W 12VDC input unit, which is quite a bit more power than I've seen in the mini-ITX plugin supply categories.
500 watts at 12VDC is 41 amps. that requires some hefty wiring, and if you have to run it any distances, either the wire is ridiculously heavy (and expensive) or you suffer from voltage drop under load.
500 watts at 120V is only 4 amps, and can easily be run 100s of feet through simple lamp cord sized wiring.
That's why the mains distribution networks use a very high AC voltage at a lower amperage. That takes care of the voltage drop across over long distances, and reduces the need for higher amperage cables.
http://en.wikipedia.org/wiki/Electric_power_distribution
Kind Regards,
Keith Roberts
----------------------------------------------------------------- Websites: http://www.karsites.net http://www.php-debuggers.net http://www.raised-from-the-dead.org.uk
All email addresses are challenge-response protected with TMDA [http://tmda.net] -----------------------------------------------------------------
John R Pierce wrote:
On 07/06/11 5:44 AM, Lamar Owen wrote:
Sure; we have a couple of small units like that for some solar-powered things we're doing here; however, the max I've seen for those plug-in type small ATX/ITX power supplies has been in the ~200W range (the specific one you linked to is only 160W), and my reply was specifically directed at the idea that lower than 48 VDC input was limited in power handling...... PowerStream has a 500W 12VDC input unit, which is quite a bit more power than I've seen in the mini-ITX plugin supply categories.
500 watts at 12VDC is 41 amps. that requires some hefty wiring, and if you have to run it any distances, either the wire is ridiculously heavy (and expensive) or you suffer from voltage drop under load.
500 watts at 120V is only 4 amps, and can easily be run 100s of feet through simple lamp cord sized wiring.
I would/will only use 12V power for direct connection from UPS battery to ATX connector. It reduces conversion losses and power draw.
Ljubomir
On 07/06/11 1:41 PM, Ljubomir Ljubojevic wrote:
I would/will only use 12V power for direct connection from UPS battery to ATX connector. It reduces conversion losses and power draw.
if by ATX connector, you mean the one on the motherboard, a standard ATX mainboard requires REGULATED 12 volts, as well as 5V and 3.3V and -5(legacy) and -12(legacy). Your 12V battery is more like 14V when its fully charged and connected to a trickle charger, and down around 11.5V when its under load and mostly depleted. If you wanted to run a mainboard on this sort of power, you would need a specially designed mainboard with built in DC-DC supplies, or you would need a DC-DC multi-voltage ATX compatible power supply (such as previously mentioned on this thread)
John R Pierce wrote:
On 07/06/11 1:41 PM, Ljubomir Ljubojevic wrote:
I would/will only use 12V power for direct connection from UPS battery to ATX connector. It reduces conversion losses and power draw.
if by ATX connector, you mean the one on the motherboard, a standard ATX mainboard requires REGULATED 12 volts, as well as 5V and 3.3V and -5(legacy) and -12(legacy). Your 12V battery is more like 14V when its fully charged and connected to a trickle charger, and down around 11.5V when its under load and mostly depleted. If you wanted to run a mainboard on this sort of power, you would need a specially designed mainboard with built in DC-DC supplies, or you would need a DC-DC multi-voltage ATX compatible power supply (such as previously mentioned on this thread)
This part of the thread is about DC input ATX power supplies, and I was referring to 12V input ATX power supply and the length of the cable between 12V source and 12V input PSU. "Direct" was meant to mean dirrectly from battery of the UPS to DC input PSU where UPS is next to the motherboard/case.
Ljubomir
On 07/06/11 2:07 PM, Ljubomir Ljubojevic wrote:
This part of the thread is about DC input ATX power supplies, and I was referring to 12V input ATX power supply and the length of the cable between 12V source and 12V input PSU. "Direct" was meant to mean dirrectly from battery of the UPS to DC input PSU where UPS is next to the motherboard/case.
ah. thats not what is commonly referred to as 'the ATX connector', so I was confused.
On Wednesday, July 06, 2011 05:23:36 PM John R Pierce wrote:
On 07/06/11 2:07 PM, Ljubomir Ljubojevic wrote:
This part of the thread is about DC input ATX power supplies,
ah. thats not what is commonly referred to as 'the ATX connector', so I was confused.
If you looked at the power supply Ljubomir previously linked to in the thread (not the PowerStream unit, but the picoPSU one), you'd see that that particular DC input power supply is built on the ATX connector itself and has no separate mechanical case. And gets 160W output power; which is excellent, for an 'on-connector' power supply. The whole supply is not much larger than the ATX connector itself; seriously, go look at this little gem.
At that point you could put a 12V power supply and a sealed lead-acid battery inside the PC case where the PSU normally goes..... you'd just have to make sure you add a schottky diode in series, since this picoPSU requires regulated 12VDC input and has overvoltage protection set around 13.0 to 13.5 volts (lead acid float voltage 13.8 typical). A 13.5 volt dry cell string and a 13.5 volt regulated power supply with a pair of 1.5V drop power diodes preventing the dry cells from charging would also work, and that sort of arrangement would indeed be a 'torch' battery (common usage here is 'flashlight' rather than 'torch').... and that would fit the needs of the OP.
The PowerStream unit can work with unregulated 9-18 volts input, and would be more suited to raw battery input. Again, a diode isolator (similar to an automotive accessory battery isolator diode set) would be required if non-rechargeable batteries were to be used as the backup.
Speaking of, I actually have some old Mirapoint rackmounts, running CentOS of course, that have built-in UPS's and redundant PSU's; haven't been able to figure out whose UPS so that I could use them with apcupsd.
Lamar Owen wrote:
On Wednesday, July 06, 2011 05:23:36 PM John R Pierce wrote:
On 07/06/11 2:07 PM, Ljubomir Ljubojevic wrote:
This part of the thread is about DC input ATX power supplies,
ah. thats not what is commonly referred to as 'the ATX connector', so I was confused.
If you looked at the power supply Ljubomir previously linked to in the thread (not the PowerStream unit, but the picoPSU one), you'd see that that particular DC input power supply is built on the ATX connector itself and has no separate mechanical case. And gets 160W output power; which is excellent, for an 'on-connector' power supply. The whole supply is not much larger than the ATX connector itself; seriously, go look at this little gem.
At that point you could put a 12V power supply and a sealed lead-acid battery inside the PC case where the PSU normally goes..... you'd just have to make sure you add a schottky diode in series, since this picoPSU requires regulated 12VDC input and has overvoltage protection set around 13.0 to 13.5 volts (lead acid float voltage 13.8 typical). A 13.5 volt dry cell string and a 13.5 volt regulated power supply with a pair of 1.5V drop power diodes preventing the dry cells from charging would also work, and that sort of arrangement would indeed be a 'torch' battery (common usage here is 'flashlight' rather than 'torch').... and that would fit the needs of the OP.
The PowerStream unit can work with unregulated 9-18 volts input, and would be more suited to raw battery input. Again, a diode isolator (similar to an automotive accessory battery isolator diode set) would be required if non-rechargeable batteries were to be used as the backup.
Speaking of, I actually have some old Mirapoint rackmounts, running CentOS of course, that have built-in UPS's and redundant PSU's; haven't been able to figure out whose UPS so that I could use them with apcupsd.
Well, it is not viable to run PC of the batteries (for long), but hooking it up directly to the battery of the UPS (so UPS charges that battery) is what I intend to do (There is nowhere to purchase them in my country yet :-( ).
Ljubomir
On Thursday, July 07, 2011 12:05:30 PM Ljubomir Ljubojevic wrote:
Well, it is not viable to run PC of the batteries (for long), but hooking it up directly to the battery of the UPS (so UPS charges that battery) is what I intend to do (There is nowhere to purchase them in my country yet :-( ).
Looking at the way the picoPSU implements the +12V output, it should be possible to use the lm_sensors package in CentOS, or the motherboard manufacturer's utility (like SuperoDoctor for Supermicro motherboards) and get an alarm and an orderly shutdown based on the +12V line's voltage. Could be an interesting application....
But until you can get one.....
On Wed, 2011-07-06 at 11:23 +0200, Ljubomir Ljubojevic wrote:
There are smaller and cheaper 12V solutions Like the picoPSU's: http://www.mini-box.com/picoPSU-160-XT
Impressive. Thanks.
Just need a 12v something to work the screen :-)
At Sat, 02 Jul 2011 14:52:27 +0200 CentOS mailing list centos@centos.org wrote:
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
Like others have suggested, a cheap UPS is the way to go.
I'm convinced. Could you (or anyone) suggest a cheap UPS? This is only a tiny server (HP MicroServer) on a home LAN.
The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and something to hold the batteries (you'll need more than one because attempting to draw a huge current from a normal battery will either kill it or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3, and several others and they need to be *precise*. Some of these are not an exact multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
Many decades ago I went to lectures at university given by Fred Hoyle (famous at the time for a TV series where he said God was unnecessary). The lectures (on thermodynamics) were not really very good, but they were interesting because Fred Hoyle was slighly paranoid, and believed evil capitalists were foisting unnecessary devices on us.
One of his pet theories was that cars did not need huge accumulators, but could be started with a torch-battery.
Another was that incandescent bulbs were deliberately made to fail after a certain time.
Another was razor blades, which according to him could easily last for ever.
One interesting idea was that instead of nuclear power stations it would be cheaper, and give the same energy, to plant trees in a strip around the equator (I forget how wide).
On Saturday 02 July 2011 15:45:11 Robert Heller wrote:
At Sat, 02 Jul 2011 14:52:27 +0200 CentOS mailing list centos@centos.org
wrote:
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and something to hold the batteries (you'll need more than one because attempting to draw a huge current from a normal battery will either kill it or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3, and several others and they need to be *precise*. Some of these are not an exact multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use only one DC battery, typically with a single voltage output, AFAIK.
Best, :-) Marko
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Marko Vojinovic Sent: Saturday, July 02, 2011 13:10 To: CentOS mailing list Subject: Re: [CentOS] Power-outage
On Saturday 02 July 2011 15:45:11 Robert Heller wrote:
At Sat, 02 Jul 2011 14:52:27 +0200 CentOS mailing list centos@centos.org
wrote:
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the
machine alive
long enough to make a graceful exit.
The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and
something to
hold the batteries (you'll need more than one because
attempting
to draw a huge current from a normal battery will
either kill it
or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
I'm sure you are right, as I know nothing at all about
power supplies.
But surely computers actually use DC, so couldn't my
torch-battery
device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3,
and several
others and they need to be *precise*. Some of these are
not an exact
multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use only one DC battery, typically with a single voltage output, AFAIK.
Best, :-) Marko
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Marko Vojinovic Sent: Saturday, July 02, 2011 13:10 To: CentOS mailing list Subject: Re: [CentOS] Power-outage
On Saturday 02 July 2011 15:45:11 Robert Heller wrote:
At Sat, 02 Jul 2011 14:52:27 +0200 CentOS mailing list centos@centos.org
wrote:
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the
machine alive
long enough to make a graceful exit.
The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and
something to
hold the batteries (you'll need more than one because
attempting
to draw a huge current from a normal battery will
either kill it
or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
I'm sure you are right, as I know nothing at all about
power supplies.
But surely computers actually use DC, so couldn't my
torch-battery
device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3,
and several
others and they need to be *precise*. Some of these are
not an exact
multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use only one DC battery, typically with a single voltage output, AFAIK.
(sorry ctrl-enter sends...)
Laptops, google mother boards, etc have power supply circuits on board. Remember that a switching powersupply taking AC still has dc to dc converters in it after the conditioning stage.
-- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - - - Jason Pyeron PD Inc. http://www.pdinc.us - - Principal Consultant 10 West 24th Street #100 - - +1 (443) 269-1555 x333 Baltimore, Maryland 21218 - - - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- This message is copyright PD Inc, subject to license 20080407P00.
On Saturday 02 July 2011 18:21:27 Jason Pyeron wrote:
But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3, and several others and they need to be *precise*. Some of these are not an exact multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use only one DC battery, typically with a single voltage output, AFAIK.
(sorry ctrl-enter sends...)
Laptops, google mother boards, etc have power supply circuits on board. Remember that a switching powersupply taking AC still has dc to dc converters in it after the conditioning stage.
So couldn't the OP then plug a battery "in between" (I'm talking in principle here, not realistically) --- after the AC-to-DC stage but before the "conditioning" stage?
If a laptop can have several *different* and *precise* voltages from a single DC battery, why the desktop cannot?
I am not saying that it would be easy or cheap, just that the above "different voltages" argument seems false from my POV. If a laptop can be battery- powered, so can a desktop (given that you have all the hardware to implement it). You don't need to tweak the motherboard, just the PSU. It's routinely done in laptops, so it doesn't seem to be rocket-science or something too expensive. I wonder why aren't there any desktops on the market with same technology?
I'm using an UPS for my desktop system, but I don't need it for the laptop. If the AC power drops, even for a moment, the laptop battery will kick in and sustain the machine. I just think that the same thing can be implemented for the desktop too. If I understood the OP correctly... ;-)
Best, :-) Marko
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org] On Behalf Of Marko Vojinovic Sent: Saturday, July 02, 2011 13:58 To: CentOS mailing list Subject: Re: [CentOS] Power-outage
On Saturday 02 July 2011 18:21:27 Jason Pyeron wrote:
But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3, and several others and they need to be *precise*. Some of
these are
not an exact multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use
only one DC
battery, typically with a single voltage output, AFAIK.
(sorry ctrl-enter sends...)
Laptops, google mother boards, etc have power supply
circuits on board.
Remember that a switching powersupply taking AC still has dc to dc converters in it after the conditioning stage.
So couldn't the OP then plug a battery "in between" (I'm talking in principle here, not realistically) --- after the AC-to-DC stage but before the "conditioning" stage?
If a laptop can have several *different* and *precise* voltages from a single DC battery, why the desktop cannot?
I am not saying that it would be easy or cheap, just that the
But the OP wanted cheap. Nor would it be easy.
above "different voltages" argument seems false from my POV. If a laptop can be battery- powered, so can a desktop (given that you have all the hardware to implement it). You don't
There is lies the rub, it is easier and cheaper to buy one, but it is even cheaper and easier to buy an UPS. The UPS comes with insurance too.
need to tweak the motherboard, just the PSU. It's routinely done in laptops, so it doesn't seem to be rocket-science or something too expensive. I wonder why aren't there any desktops on the market with same technology?
Because, it is more expensive in a world where 110 or 220 AC is standard. Our server room has DC supply for critical infrastructure, but that costs more to maintain and install. The battery room requires fire and enviromental equipment and configuration, the personnel who enter the room require certification and training on how to handle acid, hydrogen, fires, spills, etc. The electrical infrastructure required extra time and money to install because the local code did not have rules for high amprage extra low voltace (50V<low voltage<1kV) DC circuits. There are extra conduits installed, most licensed electrictions are not prepared to handle DC, so those who do cost more.........
I think you get the point.
I'm using an UPS for my desktop system, but I don't need it for the laptop. If the AC power drops, even for a moment, the
They are also low wattage, http://www.indocomp.com/ind-ups200-atx-spec.html, http://www.eurasiapower.com/power_supply_eu_eUPS-350.asp, http://www.apollo.com.tw/products_data.html
laptop battery will kick in and sustain the machine. I just think that the same thing can be implemented for the desktop too. If I understood the OP correctly... ;-)
-- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- - - - Jason Pyeron PD Inc. http://www.pdinc.us - - Principal Consultant 10 West 24th Street #100 - - +1 (443) 269-1555 x333 Baltimore, Maryland 21218 - - - -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- This message is copyright PD Inc, subject to license 20080407P00.
On 7/2/11 12:58 PM, Marko Vojinovic wrote:
If a laptop can have several *different* and *precise* voltages from a single DC battery, why the desktop cannot?
If you are facebook, you can design/demand whatever you want, including single-voltage motherboards. Not sure why everyone else put up with the problem for so long.
http://perspectives.mvdirona.com/2011/04/21/OpenComputeServerDesign.aspx
On 7/3/11, Les Mikesell lesmikesell@gmail.com wrote:
If you are facebook, you can design/demand whatever you want, including single-voltage motherboards. Not sure why everyone else put up with the problem for so long.
Legacy support and pure economies of scale :D
At Sat, 2 Jul 2011 18:58:14 +0100 CentOS mailing list centos@centos.org wrote:
On Saturday 02 July 2011 18:21:27 Jason Pyeron wrote:
But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3, and several others and they need to be *precise*. Some of these are not an exact multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use only one DC battery, typically with a single voltage output, AFAIK.
(sorry ctrl-enter sends...)
Laptops, google mother boards, etc have power supply circuits on board. Remember that a switching powersupply taking AC still has dc to dc converters in it after the conditioning stage.
So couldn't the OP then plug a battery "in between" (I'm talking in principle here, not realistically) --- after the AC-to-DC stage but before the "conditioning" stage?
The input DC side 'normal' switching power supply is like 100+ volts DC, since it is just rectified and lightly filtered from the Mains. And short of completely tearing the PC's power supply apart and rebuilding it, it is just not a trivial thing to do.
If a laptop can have several *different* and *precise* voltages from a single DC battery, why the desktop cannot?
In theory it could, it just does not need to. The AC vs. DC mains war was lost like 100 years ago. ALL main 'wall' power is AC, either 110 or 220, 50 or 60hz. Since desktops (and servers) are meant to be 'plugged into the wall', the power supplies of all want 110VAC or 220VAC (or possible either) at 50 or 60hz.
I am not saying that it would be easy or cheap, just that the above "different voltages" argument seems false from my POV. If a laptop can be battery- powered, so can a desktop (given that you have all the hardware to implement it). You don't need to tweak the motherboard, just the PSU. It's routinely done in laptops, so it doesn't seem to be rocket-science or something too expensive. I wonder why aren't there any desktops on the market with same technology?
The closest thing is a 'lunch box': a desktop system you can carry. Although once you get there, you just plug it into the nearest wall outlet. In all 'moving' situations, AC power is available: Amtrak trains all have 110V outlets at every seat. I imagine passenger aircraft also have 110V outlets ditto for ships at sea or even RVs (in both cases, hanging an alternator onto the engine is a trivial task). Building a battery powered *desktop* (or server) just does not make sense, unless you are Google (or the Army, Airforce, Navy, or NASA) and want servers that are imune to power failures, in which case you just custom build what you need.
The different voltages argument is not the argument against a battery powered desktop, just that a battery powered desktop (or server) is a non-trivial design. One cannot just solder a pair of wires onto the motherboard and connect them to some random battery. One has to start with either a motherboard *designed* to be connected to a battery (eg a laptop motherboard or Google custom server motherboard) OR a power supply meant to be connected to a DC (eg battery) power source. I believe there do exist ATX power supplies that are meant to be connected to a DC power source (like a battery). I expect that the *military* might use such in some situations.
I'm using an UPS for my desktop system, but I don't need it for the laptop. If the AC power drops, even for a moment, the laptop battery will kick in and sustain the machine. I just think that the same thing can be implemented for the desktop too. If I understood the OP correctly... ;-)
A laptop effectively contains its own UPS in the form of a power brick, battery and power supply on its motherboard.
Yes, one *could* build a desktop or server that way, but why bother, since AC wall outlets are everywhere one might want to use a desktop or server? Anyplace where a power outlet is not available, you use a laptop. A laptop is a special class (as is a smart phone or a tablet) of system. There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe, taken over progressably by laptops, tablets, smart phones, wearable computers (motherboard == shirt, monitor == shades, power supply == hat with embedded solar cells, virtual mouse/keyboard via motion sensors in your shirt sleves/gloves, etc.), or even implanted computers (eg as a thin circuit board between your skull and scalp, and 'wired' directly into your brain). This seems to already be happening to some extent, in that laptops are becoming the computer of choise and desktops are becomming an 'old school' sort of thing.
Best, :-) Marko
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Saturday 02 July 2011 21:13:59 Robert Heller wrote:
I'm using an UPS for my desktop system, but I don't need it for the laptop. If the AC power drops, even for a moment, the laptop battery will kick in and sustain the machine. I just think that the same thing can be implemented for the desktop too. If I understood the OP correctly... ;-)
A laptop effectively contains its own UPS in the form of a power brick, battery and power supply on its motherboard.
Yes, one *could* build a desktop or server that way, but why bother, since AC wall outlets are everywhere one might want to use a desktop or server?
Well, one reason I can think of is that one cannot always trust the AC wall outlet to provide uninterrupted power? By that I don't mean AC power going down for an hour, but for a fraction of a second.
I happen to live near an industrial zone, and every morning between 10 and 11h someone turns on something in the nearby factory, which makes my light-bulbs blink twice. Of course, that's enough to reboot my desktop machine no problem. I had to buy an UPS system just because of that. Granted, an UPS turned out to be a good investment, but still I wonder why there are no offers on the market with a "boosted" PSU units that can sustain DC power for a couple of seconds during the AC "blinks". They don't even need to use a battery, maybe a set of condensers could sustain power for a short period (although I might be wrong, never did the numbers on that).
A friend of mine lives near the Technical Sciences university (in the middle of the city, they don't have a campus), where they have a medium-sized wind tunnel for the aero-engineering courses. Every time they turn the thing on, the whole city block loses power for cca 5 seconds. Sure, it's bad AC grid design, but bad designs are usually a fact of life. :-)
I'm speculating here about an improved PSU device which would be more expensive than the ordinary one, but less expensive than a typical UPS system. Given that I believe there certainly is a market for such a PSU, I'm just surprised nobody is selling it yet. You cannot assume that the AC wall outlet will always provide perfect power... ;-)
There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe, taken over progressably by laptops, tablets, smart phones, wearable computers (motherboard == shirt, monitor == shades, power supply == hat with embedded solar cells, virtual mouse/keyboard via motion sensors in your shirt sleves/gloves, etc.),
I could in principle imagine all that coming in the future, but the "monitor == shades" thing is just only Fi with no Sci in it. A human eye cannot focus properly on any object which is closer to the eye than 10-15 cm (depending on the eye quality), so there is absolutely no way one can use shades or contact lenses or something similar as a monitor, regardless of technological levels of any human or alien races (James Bond notwithstanding). Unless of course one surgically adapts the eye lense itself, in which case the person would not be able to see anything else... ;-)
or even implanted computers (eg as a thin circuit board between your skull and scalp, and 'wired' directly into your brain).
I would never wire a brain to a machine. Brains make errors, are susceptible to emotions, hormons, vanity, etc., and just introduce a large point of failure for the otherwise-correct machines. ;-)
This seems to already be happening to some extent, in that laptops are becoming the computer of choise and desktops are becomming an 'old school' sort of thing.
Yeah, the laptops are becoming cheap enough, so that once your computing needs grow out of your current laptop, you don't even think of "upgrading" it, but rather just buy a new model. Desktops will be in use only for custom things (professionals who need, say, five audio cards in one machine) and small servers. But we're getting sort-of OT here... ;-)
Best, :-) Marko
At Sun, 3 Jul 2011 00:34:18 +0100 CentOS mailing list centos@centos.org wrote:
On Saturday 02 July 2011 21:13:59 Robert Heller wrote:
I'm using an UPS for my desktop system, but I don't need it for the laptop. If the AC power drops, even for a moment, the laptop battery will kick in and sustain the machine. I just think that the same thing can be implemented for the desktop too. If I understood the OP correctly... ;-)
A laptop effectively contains its own UPS in the form of a power brick, battery and power supply on its motherboard.
Yes, one *could* build a desktop or server that way, but why bother, since AC wall outlets are everywhere one might want to use a desktop or server?
Well, one reason I can think of is that one cannot always trust the AC wall outlet to provide uninterrupted power? By that I don't mean AC power going down for an hour, but for a fraction of a second.
I happen to live near an industrial zone, and every morning between 10 and 11h someone turns on something in the nearby factory, which makes my light-bulbs blink twice. Of course, that's enough to reboot my desktop machine no problem. I had to buy an UPS system just because of that. Granted, an UPS turned out to be a good investment, but still I wonder why there are no offers on the market with a "boosted" PSU units that can sustain DC power for a couple of seconds during the AC "blinks". They don't even need to use a battery, maybe a set of condensers could sustain power for a short period (although I might be wrong, never did the numbers on that).
A friend of mine lives near the Technical Sciences university (in the middle of the city, they don't have a campus), where they have a medium-sized wind tunnel for the aero-engineering courses. Every time they turn the thing on, the whole city block loses power for cca 5 seconds. Sure, it's bad AC grid design, but bad designs are usually a fact of life. :-)
I'm speculating here about an improved PSU device which would be more expensive than the ordinary one, but less expensive than a typical UPS system. Given that I believe there certainly is a market for such a PSU, I'm just surprised nobody is selling it yet. You cannot assume that the AC wall outlet will always provide perfect power... ;-)
There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe, taken over progressably by laptops, tablets, smart phones, wearable computers (motherboard == shirt, monitor == shades, power supply == hat with embedded solar cells, virtual mouse/keyboard via motion sensors in your shirt sleves/gloves, etc.),
I could in principle imagine all that coming in the future, but the "monitor == shades" thing is just only Fi with no Sci in it. A human eye cannot focus properly on any object which is closer to the eye than 10-15 cm (depending on the eye quality), so there is absolutely no way one can use shades or contact lenses or something similar as a monitor, regardless of technological levels of any human or alien races (James Bond notwithstanding). Unless of course one surgically adapts the eye lense itself, in which case the person would not be able to see anything else... ;-)
Hmmm... There were a CS prof. and some students at UMass when I was working there playing with a computer in a backpack with a 1" monitor suspended from a head mount in front of one eye. Not anything like 10-15 cm. If 10-15 cm is the minimum distance, what about telescope eyepieces, camera viewfinders (including the little video ones on camcorders), or binoculars? *I* know I can see images in the video viewfinder of my Sony Hi8 camcorder just fine, with my right up close (the old camcorder I have does NOT have a 3" swing out monitor). It is all about the optics.
or even implanted computers (eg as a thin circuit board between your skull and scalp, and 'wired' directly into your brain).
I would never wire a brain to a machine. Brains make errors, are susceptible to emotions, hormons, vanity, etc., and just introduce a large point of failure for the otherwise-correct machines. ;-)
This seems to already be happening to some extent, in that laptops are becoming the computer of choise and desktops are becomming an 'old school' sort of thing.
Yeah, the laptops are becoming cheap enough, so that once your computing needs grow out of your current laptop, you don't even think of "upgrading" it, but rather just buy a new model. Desktops will be in use only for custom things (professionals who need, say, five audio cards in one machine) and small servers. But we're getting sort-of OT here... ;-)
Best, :-) Marko
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Sunday 03 July 2011 00:51:29 Robert Heller wrote:
There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe, taken over progressably by laptops, tablets, smart phones, wearable computers (motherboard == shirt, monitor == shades, power supply == hat with embedded solar cells, virtual mouse/keyboard via motion sensors in your shirt sleves/gloves, etc.),
I could in principle imagine all that coming in the future, but the "monitor == shades" thing is just only Fi with no Sci in it. A human eye cannot focus properly on any object which is closer to the eye than 10-15 cm (depending on the eye quality), so there is absolutely no way one can use shades or contact lenses or something similar as a monitor, regardless of technological levels of any human or alien races (James Bond notwithstanding). Unless of course one surgically adapts the eye lense itself, in which case the person would not be able to see anything else... ;-)
Hmmm... There were a CS prof. and some students at UMass when I was working there playing with a computer in a backpack with a 1" monitor suspended from a head mount in front of one eye. Not anything like 10-15 cm. If 10-15 cm is the minimum distance, what about telescope eyepieces, camera viewfinders (including the little video ones on camcorders), or binoculars? *I* know I can see images in the video viewfinder of my Sony Hi8 camcorder just fine, with my right up close (the old camcorder I have does NOT have a 3" swing out monitor). It is all about the optics.
I wouldn't know about that CS prof. at UMass. Have any info that can point me to him? Other examples you mention all have to do with lenses that twist the trajectory of light to make distant or small things visible. When using telescopes, binoculars, camera viewfinders, microscopes, and other stuff like that, you are actually looking *through* a (transparent) device to see something else outside, you're never looking *at* a device, or something that is inside it.
In contrast to that, actually drawing a picture which is 1-2cm away from the eye is a completely different game. Just take a piece of paper, draw something on it and put it 2 cm in front of your eye. The drawing will get blurred. And it's not because you used a thick pen, but because the eye lens cannot focus on such a short distance.
Now, you might consider putting some convenient lenses between the paper and the eye, to fix that problem. I don't have time do actually do the calculation of the properties of such a lens, but it's an interesting problem in geometric optics. You would want a convex lens that moves the focal point of the eye from 15 cm to 2 cm. The trick is to find a transparent material which would have a refraction index high enough that it can do what you want, while still be thin enough to fit between the monitor and the eye (ie. it needs to be thinner than 2 cm). I don't know if ordinary glass or any other material would do that or not. But it could be an interesting exercise for a student of geometric optics. :-)
The bigger issue is the fact that, even if you manage to find an appropriate lens to move the focal point to 2 cm, it is going to distort everything else you see behind it. In principle you could devote one eye for the monitor-only, making the whole apparatus non-transparent, and use the other eye for the outside world. That would, however, destroy the 3D vision of both the outside world and eventual monitor 3D picture (because you can wear it only on one eye).
Actually, now that I think more and more about it, I am not so sure it is not doable. However, it is far from being trivial, and it certainly cannot be something that can be as thin as ordinary shades. It has to be bulky and heavy (due to the optics inside) and is bound to impair your vision of the real world.
If I get some free time, I might even try to calculate the properties of such a system of lenses, but I'm skeptic that the cool "monitor-shades" will ever be possible. ;-)
But now we are getting quite OT here... ;-)
Best, :-) Marko
At Sun, 3 Jul 2011 04:30:37 +0100 CentOS mailing list centos@centos.org wrote:
On Sunday 03 July 2011 00:51:29 Robert Heller wrote:
There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe, taken over progressably by laptops, tablets, smart phones, wearable computers (motherboard == shirt, monitor == shades, power supply == hat with embedded solar cells, virtual mouse/keyboard via motion sensors in your shirt sleves/gloves, etc.),
I could in principle imagine all that coming in the future, but the "monitor == shades" thing is just only Fi with no Sci in it. A human eye cannot focus properly on any object which is closer to the eye than 10-15 cm (depending on the eye quality), so there is absolutely no way one can use shades or contact lenses or something similar as a monitor, regardless of technological levels of any human or alien races (James Bond notwithstanding). Unless of course one surgically adapts the eye lense itself, in which case the person would not be able to see anything else... ;-)
Hmmm... There were a CS prof. and some students at UMass when I was working there playing with a computer in a backpack with a 1" monitor suspended from a head mount in front of one eye. Not anything like 10-15 cm. If 10-15 cm is the minimum distance, what about telescope eyepieces, camera viewfinders (including the little video ones on camcorders), or binoculars? *I* know I can see images in the video viewfinder of my Sony Hi8 camcorder just fine, with my right up close (the old camcorder I have does NOT have a 3" swing out monitor). It is all about the optics.
I wouldn't know about that CS prof. at UMass. Have any info that can point me to him? Other examples you mention all have to do with lenses that twist the trajectory of light to make distant or small things visible. When using telescopes, binoculars, camera viewfinders, microscopes, and other stuff like that, you are actually looking *through* a (transparent) device to see something else outside, you're never looking *at* a device, or something that is inside it.
I don't remember who was doing the experiements.
It would likely be a 1" camcorder viewfinder 'monitor', that is designed to be right up against one's eye.
In contrast to that, actually drawing a picture which is 1-2cm away from the eye is a completely different game. Just take a piece of paper, draw something on it and put it 2 cm in front of your eye. The drawing will get blurred. And it's not because you used a thick pen, but because the eye lens cannot focus on such a short distance.
Now, you might consider putting some convenient lenses between the paper and the eye, to fix that problem. I don't have time do actually do the calculation of the properties of such a lens, but it's an interesting problem in geometric optics. You would want a convex lens that moves the focal point of the eye from 15 cm to 2 cm. The trick is to find a transparent material which would have a refraction index high enough that it can do what you want, while still be thin enough to fit between the monitor and the eye (ie. it needs to be thinner than 2 cm). I don't know if ordinary glass or any other material would do that or not. But it could be an interesting exercise for a student of geometric optics. :-)
The bigger issue is the fact that, even if you manage to find an appropriate lens to move the focal point to 2 cm, it is going to distort everything else you see behind it. In principle you could devote one eye for the monitor-only, making the whole apparatus non-transparent, and use the other eye for the outside world. That would, however, destroy the 3D vision of both the outside world and eventual monitor 3D picture (because you can wear it only on one eye).
Actually, now that I think more and more about it, I am not so sure it is not doable. However, it is far from being trivial, and it certainly cannot be something that can be as thin as ordinary shades. It has to be bulky and heavy (due to the optics inside) and is bound to impair your vision of the real world.
If I get some free time, I might even try to calculate the properties of such a system of lenses, but I'm skeptic that the cool "monitor-shades" will ever be possible. ;-)
OTOH, I would expect that people in the 1890's would consider the Apollo Moon landings as 'impossible'... So, given enough advances in optics and monitor techology: LCD screens that can switch to complete transparency or to varying levels of transparency, and things like programmable lenses / optical systems, it becomes concievable.
But now we are getting quite OT here... ;-)
Best, :-) Marko
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Sunday 03 July 2011 13:21:31 Robert Heller wrote:
Actually, now that I think more and more about it, I am not so sure it is not doable. However, it is far from being trivial, and it certainly cannot be something that can be as thin as ordinary shades. It has to be bulky and heavy (due to the optics inside) and is bound to impair your vision of the real world.
If I get some free time, I might even try to calculate the properties of such a system of lenses, but I'm skeptic that the cool "monitor-shades" will ever be possible. ;-)
OTOH, I would expect that people in the 1890's would consider the Apollo Moon landings as 'impossible'... So, given enough advances in optics and monitor techology: LCD screens that can switch to complete transparency or to varying levels of transparency, and things like programmable lenses / optical systems, it becomes concievable.
Well, sure, but you have to make a very fine distinction between prejudices and facts, when it comes to making these sorts of predictions. I remember to have read somewhere that in 18th century it was widely believed that the human body cannot withstand velocities greater than 60 km/h, for whatever reason. But this was just due to the people's ignorance of the difference between velocity and acceleration. But when it comes to making claims about the optical properties of the human eye, no prejudices are involved --- it has been well explored and studied, and limits of what the eye can see (without artificial help) are a matter of fact, rather than unsupported opinion. ;-)
There is *no* *way* to have a plain eye read anything at 2 cm distance --- this is a fact, easily verifiable if you put your palm in front of your eye. Now, with some clever lenses and optics you might be able to trick the eye into thinking that the object is at a greater distance, and therefore circumvent this limitation. But having these optics constrained onto a width of typical shades is an *extreme* stretch, ultimately bounded by laws of physics. The laws of physics that apply here are already known, well-studied laws of geometric optics, and have to be obeyed regardless of any technological ingenuity. I haven't done any serious research, but I can say offhand that they provide very very little space for any hi-tech manouvers, no matter how advanced.
The more likely approach to what you want is to have some apparatus that dynamically measures the current width of the eye lens, and use a super- tightly focused laser to draw the monitor picture directly on the eye retina. AFAIK, that is concievable, *provided* you can create a laser beam thin enough not to disperse drastically going through the eye lens, and thin enough to provide good "monitor resolution", and make it in appropriate color(s), and have low-enough power not to damage the retina on continuous exposure. By dynamically measuring the eye refraction properties, the device could in principle adjust the laser drawing to accomodate to the current position and the focus of the eye itself. The end result would be a person seeing glowing text "written" on the object that the eyes are currently looking at. When you look at something else, the device adjusts and you see the text written over a new object.
*If* you are technologically advanced enough to create a laser beam with such properties, this kind of a thing would be doable. Note, though, that the requrements of the beam being thin enough and low-enough in power may be in conflict with each other, and not satisfiable simultaneously (to be honest, I'm not sure without some digging into it, but at the very end you cannot circumvent Heisenberg's uncertainty principle, if all else can be done). But if that can be done, I can imagine a device like that to be constructed somehow.
Nevertheless, even then, this device would look like two lasers pointed into your eyes, and certainly *not* like cool shades. ;-)
Best, :-) Marko
On 7/3/11, Marko Vojinovic vvmarko@gmail.com wrote:
be a good investment, but still I wonder why there are no offers on the market with a "boosted" PSU units that can sustain DC power for a couple of seconds during the AC "blinks". They don't even need to use a battery, maybe a set of condensers could sustain power for a short period (although I might be wrong, never did the numbers on that).
Given that I believe there certainly is a market for such a PSU, I'm just surprised nobody is selling it yet. You cannot assume that the AC wall outlet will always provide perfect power... ;-)
Ah, manufacturers do sell this stuff at least a few years back. It came with a battery pack that was installed into the 5.25" drive bay. But as this type never made it into mainstream awareness, likely due to the low cost of a basic UPS nowadays, it just didn't make sense when such a PSU cost almost the same as a normal PSU + UPS. And the UPS had a longer run time as well as being able to support a few other accessories at the same time, and acts as a shield by sitting between any power surge and your PSU, as opposed to the PSU + battery.
On 7/2/2011 7:34 PM, Marko Vojinovic wrote:
I could in principle imagine all that coming in the future, but the "monitor == shades" thing is just only Fi with no Sci in it. A human eye cannot focus properly on any object which is closer to the eye than 10-15 cm (depending on the eye quality), so there is absolutely no way one can use shades or contact lenses or something similar as a monitor, regardless of technological levels of any human or alien races (James Bond notwithstanding). Unless of course one surgically adapts the eye lense itself, in which case the person would not be able to see anything else... ;-)
Hmm...something like this perhaps?
http://www.i-glassesstore.com/i-3d.html
Still a bit bulky and expensive, but not impossible. These apparently use a lens of some sort to allow the eye to focus at 5' while wearing them. I had the chance to play with a pair of these 10 years ago. At that time, the resolution sucked and they were about 1.5" thick. They had built-in motion tracking. Playing Descent with those things was a blast! :)
On Tuesday 05 July 2011 21:31:50 Bowie Bailey wrote:
On 7/2/2011 7:34 PM, Marko Vojinovic wrote:
I could in principle imagine all that coming in the future, but the "monitor == shades" thing is just only Fi with no Sci in it. A human eye cannot focus properly on any object which is closer to the eye than 10-15 cm (depending on the eye quality), so there is absolutely no way one can use shades or contact lenses or something similar as a monitor, regardless of technological levels of any human or alien races (James Bond notwithstanding). Unless of course one surgically adapts the eye lense itself, in which case the person would not be able to see anything else... ;-)
Hmm...something like this perhaps?
http://www.i-glassesstore.com/i-3d.html
Still a bit bulky and expensive, but not impossible. These apparently use a lens of some sort to allow the eye to focus at 5' while wearing them. I had the chance to play with a pair of these 10 years ago. At that time, the resolution sucked and they were about 1.5" thick. They had built-in motion tracking. Playing Descent with those things was a blast! :)
Well, yes, sure, but they have to have this lens to allow the eye to focus properly, as you said. This means two crucial things: (a) they have to be bulky, and (b) there is no way to make them transparent, so that a person can watch the monitor picture and the outside world simultaneously.
By (b) I mean having computer graphics overlayed on top of real-world scenery (like in Terminator or Robocop movies). I'm just saying that this kind of overlay is impossible to achieve with a regular human eye, except with very bulky equipment hanging off your head 15 cm in front of your face.
Anyway, for both of the above reasons, these AV-headsets don't qualify even as a predecessor of "monitor-shades" (which are by assumption thin and transparent). These things are used for virtual reality applications, and they have existed for some time now. But as long as they cut you off from real world, I cannot consider them as "shades" in any way. ;-)
Best, :-) Marko
On 7/6/11, Marko Vojinovic vvmarko@gmail.com wrote:
By (b) I mean having computer graphics overlayed on top of real-world scenery (like in Terminator or Robocop movies). I'm just saying that this kind of overlay is impossible to achieve with a regular human eye, except with very bulky equipment hanging off your head 15 cm in front of your face.
Somehow it just doesn't seem impossible to me, unless we insist on a certain form-factor for the glasses as opposed to the sole requirement that the frontal portion should be relatively thin.
How about a glass with lenses that are formed by nano beam splitter/mirrors with projection units on the side where the legs of the glasses would be. Thus transferring the bulk from the front to the sides which would be more wearable than a heavy weight hanging off the nose. The overlay would be generated by the projection units which is reflected into the eye via the front mirror/beamsplitter which would also allow external light in, albeit at half strength but that would be what "shades" do after all ;)
Greetings,,
On 7/3/11, Robert Heller heller@deepsoft.com wrote:
At Sat, 2 Jul 2011 18:58:14 +0100 CentOS mailing list centos@centos.org wrote: There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe,
Something like this?: http://www.youtube.com/watch?v=YrtANPtnhyg
Rajagopal Swaminathan wrote:
Greetings,,
On 7/3/11, Robert Heller heller@deepsoft.com wrote:
At Sat, 2 Jul 2011 18:58:14 +0100 CentOS mailing list centos@centos.org wrote: There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe,
Something like this?: http://www.youtube.com/watch?v=YrtANPtnhyg
Amassing technology synergy.
I liked even more his resolve to offer this as open source.
Ljubomir
Ljubomir Ljubojevic wrote:
Rajagopal Swaminathan wrote:
Greetings,,
On 7/3/11, Robert Heller heller@deepsoft.com wrote:
At Sat, 2 Jul 2011 18:58:14 +0100 CentOS mailing list centos@centos.org wrote: There is (in the SciFi world) the idea that someday 'desktops' in the current / conventional sense may completely vanish from the universe,
Something like this?: http://www.youtube.com/watch?v=YrtANPtnhyg
Amassing technology synergy.
I liked even more his resolve to offer this as open source.
Hmmm, no open source for 3 years now. I might need to take my praise back.
Ljubomir
At Sat, 2 Jul 2011 18:10:09 +0100 CentOS mailing list centos@centos.org wrote:
On Saturday 02 July 2011 15:45:11 Robert Heller wrote:
At Sat, 02 Jul 2011 14:52:27 +0200 CentOS mailing list centos@centos.org
wrote:
Emmanuel Noobadmin wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit.
The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and something to hold the batteries (you'll need more than one because attempting to draw a huge current from a normal battery will either kill it or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
I'm sure you are right, as I know nothing at all about power supplies. But surely computers actually use DC, so couldn't my torch-battery device just supply the PC components directly?
A PC uses several *different* DC voltages: +12, +5, +3.3, and several others and they need to be *precise*. Some of these are not an exact multiples of the standard 1.5V Carbon-Zinc cells typicaly used in torch batteries.
I wonder, how is this issue solved in laptops? They use only one DC battery, typically with a single voltage output, AFAIK.
*Mobile* processors don't use as many voltages. You will note that laptop disks use a single 5V for power, as opposed to the 5V (logic) and 12V (motor) power a desktop or server disk drive uses.
Laptops also use DC-DC power converters to get the addition *few* volatages needed (eg 12 => 5V and 3.3V).
(I'm guessing that Google's massive array of battery powered servers use things like Atom processors or some other mobile or embeded processing elements on their custom motherboards, which might be a customized variant of a laptop or embedded processing motherboard.)
Best, :-) Marko
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 7/3/11, Robert Heller heller@deepsoft.com wrote:
(I'm guessing that Google's massive array of battery powered servers use things like Atom processors or some other mobile or embeded processing elements on their custom motherboards, which might be a customized variant of a laptop or embedded processing motherboard.)
IIRC, Google's server boards run normal processors. It's just the board was customized to work off single+12V voltage, with a battery sitting in between to act as a UPS. The reason was a single +12V battery per server was cheaper than a huge UPS for a bunch of servers. Google also use enough boards that it was economically feasible for the manufacturer to make design changes to an existing board design to support this.
On Sat, 2011-07-02 at 03:03 +0800, Emmanuel Noobadmin wrote:
On 7/1/11, Timothy Murphy gayleard@eircom.net wrote:
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
Like others have suggested, a cheap UPS is the way to go. The problem with your idea is that you'll need a DC to AC inverter that can handle the output current required by your server and something to hold the batteries (you'll need more than one because attempting to draw a huge current from a normal battery will either kill it or at the very least cause it to have a shorter than expected capacity) and everything together, it's probably going to cost more in both money and time to have this thing.
You will also need to have the device signal the OS to shutdown cleanly and be set to reboot when the power comes back on.
And once you've added those features, you will have created a UPS -- likely at an expense in time/money that exceeds simply having bought one. Specifically, a 300W UPS can be had for less than $40 -- that's 0.5~4 hours of overtime or side work depending on your job. You are likely to expend a lot more than 4 hours putting your homemade solution together and achieve a far less reliable result. Unless the experience of amateur electrical engineering is what you are craving (it *is* fun) buy a UPS and be done with it -- just read the docs so you fully understand how to make it tell your computer to start shutting down or booting up, etc. They aren't magic and require a little set up to be fully utilized.
You should invest in a Spider KVM or similar, they hang off the back and don't use any rack space. They can also be POE, so they wont use a plug. That'll provide you out of band management and remote reboots and what not.
On Fri, Jul 1, 2011 at 7:05 AM, Timothy Murphy gayleard@eircom.net wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
My server, an HP MicroServer, came back (re-booted) on 2 of the 3 occasions, but not on the third.
I assume that the problem arises because the machine does not close down properly. (Although it is also possible that a voltage surge might have been responsible - I have no surge protector on this supply.)
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
I know there is a Remote Management (iLO) card for this machine, which might be useful for this. Unfortunately, I've already used the PCIe slot for a second ethernet card.
Any advice or suggestions gratefully received.
-- Timothy Murphy e-mail: gayleard /at/ eircom.net tel: +353-86-2336090, +353-1-2842366 s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Steven Crothers wrote:
You should invest in a Spider KVM or similar, they hang off the back and don't use any rack space. They can also be POE, so they wont use a plug. That'll provide you out of band management and remote reboots and what not.
How exactly would that work?
Hook up ethernet, if its not POE, you plug it in, attach all the various usb cables, vga, serial, ps/2, ect ect to the server and let it hang. When your server is unresponsive just go ahead and hit the IP you assigned to your Spider, and you get a full console, virtual media, mass storage emulation, and the ability to mount samba shares and what not into it.
On Sun, Jul 3, 2011 at 3:00 AM, Timothy Murphy gayleard@eircom.net wrote:
Steven Crothers wrote:
You should invest in a Spider KVM or similar, they hang off the back and don't use any rack space. They can also be POE, so they wont use a plug. That'll provide you out of band management and remote reboots and what not.
How exactly would that work?
-- Timothy Murphy e-mail: gayleard /at/ eircom.net tel: +353-86-2336090, +353-1-2842366 s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Steven Crothers wrote:
Hook up ethernet, if its not POE, you plug it in, attach all the various usb cables, vga, serial, ps/2, ect ect to the server and let it hang. When your server is unresponsive just go ahead and hit the IP you assigned to your Spider, and you get a full console, virtual media, mass storage emulation, and the ability to mount samba shares and what not into it.
How exactly would that work?
I'm still not clear on this solution. Assuming you are actually doing this, could you tell me how you set it up in a little more detail, please.
Timothy Murphy wrote:
Steven Crothers wrote:
Hook up ethernet, if its not POE, you plug it in, attach all the various usb cables, vga, serial, ps/2, ect ect to the server and let it hang. When your server is unresponsive just go ahead and hit the IP you assigned to your Spider, and you get a full console, virtual media, mass storage emulation, and the ability to mount samba shares and what not into it.
How exactly would that work?
I'm still not clear on this solution. Assuming you are actually doing this, could you tell me how you set it up in a little more detail, please.
You hook up device to the PC, and both to internet, device with public IP, best if it is static, or with dynamic domain.
Then you use (app or web browser?) and open up IP of the device and you get somethink like VNC or TeamViewer but directly to hardware. Device has some sort of embedded OS in the firmware so you have access to all of your data, or at least I understood his comment in that way. Anyway, you can remotely even access BIOS screen.
Ljubomir
Ljubomir Ljubojevic wrote:
Hook up ethernet, if its not POE, you plug it in, attach all the various usb cables, vga, serial, ps/2, ect ect to the server and let it hang. When your server is unresponsive just go ahead and hit the IP you assigned to your Spider, and you get a full console, virtual media, mass storage emulation, and the ability to mount samba shares and what not into it.
How exactly would that work?
I'm still not clear on this solution. Assuming you are actually doing this, could you tell me how you set it up in a little more detail, please.
You hook up device to the PC, and both to internet, device with public IP, best if it is static, or with dynamic domain.
I'm not sure how I would connect both Spider and PC to the internet. If I had to purchase a second IP address, and pay my ISP for a second line, with the cost of the Spider this would be getting quite expensive. Maybe I have misunderstood something, as I am no network guru.
Timothy Murphy wrote:
Ljubomir Ljubojevic wrote:
Hook up ethernet, if its not POE, you plug it in, attach all the various usb cables, vga, serial, ps/2, ect ect to the server and let it hang. When your server is unresponsive just go ahead and hit the IP you assigned to your Spider, and you get a full console, virtual media, mass storage emulation, and the ability to mount samba shares and what not into it.
How exactly would that work?
I'm still not clear on this solution. Assuming you are actually doing this, could you tell me how you set it up in a little more detail, please.
You hook up device to the PC, and both to internet, device with public IP, best if it is static, or with dynamic domain.
I'm not sure how I would connect both Spider and PC to the internet. If I had to purchase a second IP address, and pay my ISP for a second line, with the cost of the Spider this would be getting quite expensive. Maybe I have misunderstood something, as I am no network guru.
You said your server is not directly on the internet, so I guess you have some sort of the router/firewall. On the router, direct (DNAT) needed ports to KVM and the rest to the Server.
Ljubomir
They have cheaper smaller UPS's that should be able to help you.
On Fri, Jul 1, 2011 at 7:05 AM, Timothy Murphy gayleard@eircom.net wrote:
I have a CentOS-5.6 remote server in a house in Italy, where there are occasional thunder-storms.
There was one yesterday, when the electricity went off 3 times, for a second or so on each occasion.
My server, an HP MicroServer, came back (re-booted) on 2 of the 3 occasions, but not on the third.
I assume that the problem arises because the machine does not close down properly. (Although it is also possible that a voltage surge might have been responsible - I have no surge protector on this supply.)
It seems to me that it should be possible to have a simple, torch-battery operated, system which will keep the machine alive long enough to make a graceful exit. A full-blown UPS would be excessive, I think, as I only want the machine to re-boot when the current comes back on.
I know there is a Remote Management (iLO) card for this machine, which might be useful for this. Unfortunately, I've already used the PCIe slot for a second ethernet card.
Any advice or suggestions gratefully received.
-- Timothy Murphy e-mail: gayleard /at/ eircom.net tel: +353-86-2336090, +353-1-2842366 s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
James Matthews wrote:
They have cheaper smaller UPS's that should be able to help you.
What UPS's are you suggesting? (I didn't really follow your remark.)
On 07/01/11 4:05 AM, Timothy Murphy wrote:
It seems to me that it should be possible to have a simple, torch-battery operated
assuming 'torch' in this context means what us yank's call a flashlight, and that a 'torch battery' is a C or D cell, lets see how much juice we could get out of a reasonable setup..
According to HP's 'quick specs', the Microserver has a 150W supply, that nominally draws 0.63 A (from 115VAC) or 0.35 A rom 220VAC, about 75 watts.
According to wikipedia, an alkaline D cell is typically rated at 1.2 amp hours at 1.5 volts. Most inexpensive AC inverters run on 12V (automobile power), so we'd need 8 of them to get 1.2AH at 12V or about 14 watt*hours... My calculations seem to suggest you'd get maybe 9 minutes total from those 8 batteries with an 80% efficient AC inverter at 75 watts. And then of course, you'd be throwing those 8 D cells away and replacing them with new ones.
NiCAD or NiMH rechargeable batteries aren't particularly suitable for this application, you'd need 10 of them as they are 1.2V, and they don't do well as standby power since they self discharge when idle for long periods. UPS's almost always use lead-acid batteries as they are far more suitable for standby power applications.
A UPS is little more than a rechargeable battery, an A/C inverter circuit, a battery charger, and a controller for all that which also signals your computer when the power is failing.
John R Pierce wrote:
On 07/01/11 4:05 AM, Timothy Murphy wrote:
It seems to me that it should be possible to have a simple, torch-battery operated
assuming 'torch' in this context means what us yank's call a flashlight, and that a 'torch battery' is a C or D cell, lets see how much juice we could get out of a reasonable setup..
According to HP's 'quick specs', the Microserver has a 150W supply, that nominally draws 0.63 A (from 115VAC) or 0.35 A rom 220VAC, about 75 watts.
In fact, my MicroServer has never exceeded 45 watts, and usually runs at about 30 watts. I'm not sure what sort of current a computer uses when shutting down?
According to wikipedia, an alkaline D cell is typically rated at 1.2 amp hours at 1.5 volts. Most inexpensive AC inverters run on 12V (automobile power), so we'd need 8 of them to get 1.2AH at 12V or about 14 watt*hours... My calculations seem to suggest you'd get maybe 9 minutes total from those 8 batteries with an 80% efficient AC inverter at 75 watts. And then of course, you'd be throwing those 8 D cells away and replacing them with new ones.
Actually, I would only want the batteries to last long enough to shut down the computer cleanly, which I think takes about 30 seconds. And we can get quite large 12 volt batteries pretty cheaply this side of the pond.
Nb I didn't say, or mean to say, that I wanted to _make_ a "flashlight UPS". My electronic skill is close to zero. I was simply expressing surprise that no-one had done it.
I've been completely convinced that a UPS is what I need, and am trying to source the APC UPS-BE350G, which was recommended.
On Tue, Jul 5, 2011 at 12:14 AM, Timothy Murphy gayleard@eircom.net wrote:
John R Pierce wrote:
Nb I didn't say, or mean to say, that I wanted to _make_ a "flashlight UPS". My electronic skill is close to zero.
Well, that's your first problem - not knowing how todo it :)
I was simply expressing surprise that no-one had done it.
I have already done it. Years ago. In fact, my home server can run 2 hours on batteries, and my ADSL modem + 24port GB switch can run about 14hours on batteries.
I've been completely convinced that a UPS is what I need, and am trying to source the APC UPS-BE350G, which was recommended.
In your case the UPS is the easiest route to follow
--
Timothy Murphy wrote:
I've been completely convinced that a UPS is what I need, and am trying to source the APC UPS-BE350G, which was recommended.
I used APC Back-UPS CS BK500EI in a company that I service and with their app (on Windows) I stretched lower and higher Volt boundary to something close to 165-240V (for 220V system) and they successfully stand cheaper emergency power generator when power goes out for a longer period of time.
Ljubomir
On 7/4/11 5:14 PM, Timothy Murphy wrote:
I've been completely convinced that a UPS is what I need, and am trying to source the APC UPS-BE350G, which was recommended.
One thing that might not have been mentioned yet: somewhere around three years out, a small UPS will cause an outage you wouldn't have otherwise when it fails. Generally replacing the battery will fix this, but by 6-10 years other components will go too.
On 07/04/11 3:40 PM, Les Mikesell wrote:
One thing that might not have been mentioned yet: somewhere around three years out, a small UPS will cause an outage you wouldn't have otherwise when it fails. Generally replacing the battery will fix this, but by 6-10 years other components will go too.
indeed, about half the times I've replaced the battery in a APC BackUPS or other low end UPS, the UPS itself was dead.
Greetings,
On 7/5/11, Les Mikesell lesmikesell@gmail.com wrote:
On 7/4/11 5:14 PM, Timothy Murphy wrote:
Generally replacing the battery will fix this, but by 6-10 years other components will go too.
<rant> I was running a shop with two servers as ltsp with about 100 thin clients and a dozen projectors.
One 20 KVA UPS powered all of them
There was another 25KVS for critical fan light etc.
Withing two years at least 20 (out of 64 IIRC) ("mainance free") batteries were replaced.
I insisted on water test on the batteries.
I would suggest do a water on test batteries _before_ buying/installing them
The entire battery pack was replaced. Thank god they were under warranty. Otherwise I would have been skinless probably.
beware of maitenance free ones.
As the one man show , of course I will not mention the woe/sob/curses Ihad to face due to outages. An oh almost all the users werere PHBs/ or would be PHBs </rant>
On 7/5/11, Rajagopal Swaminathan raju.rajsand@gmail.com wrote:
<rant> I was running a shop with two servers as ltsp with about 100 thin clients and a dozen projectors.
One 20 KVA UPS powered all of them
There was another 25KVS for critical fan light etc.
Withing two years at least 20 (out of 64 IIRC) ("mainance free") batteries were replaced.
Fortunately I read about Google's server and their cheap per server UPS approach and never went for the single high power UPS except when absolutely needed. It was easier to deal with dead UPS in places where there are more than one server simply by plugging the machine into the other UPS while the first was replaced.
Otherwise, the single large UPS becomes the single point of failure.
On 07/04/11 11:44 PM, Emmanuel Noobadmin wrote:
Otherwise, the single large UPS becomes the single point of failure.
the good big ones are fully redundant and every component is hot swappable.
but yeah, distributed UPS the way google did it is rather sweet. As long as part of their operating plan is replacing every server by the time its batteries need replacing, they are good to go. You also need the sort of cloud management environment where your servers themselves are interchangable worker units and can be easily taken down for maintenance
OTOH, populating lots of racks with several APC SU3000's each is a pain in the butt. Servicing dozens or 100s of per-rack UPS's when they are 3-4 years old and need battery replacement is also very time consuming. a battery tray swap of a decent sized datacenter UPS takes a half day with a small fork lift.
On Tue, Jul 05, 2011 at 02:44:30PM +0800, Emmanuel Noobadmin wrote:
Can people at least pretend to keep this list on-topic? 89 responses for an off-topic post is a little much, don't you think?
Item 3 under Guidelines as listed at:
http://www.centos.org/modules/tinycontent/index.php?id=16
John
-
Always Learning -
Blake Hudson -
Bowie Bailey -
Brunner, Brian T. -
Christopher Chan -
Colin Coles -
Devin Reade -
Drew -
Emmanuel Noobadmin -
Fajar Priyanto -
fred smith -
James Matthews -
Jason Pyeron -
Jim Nelson -
John Hodrien -
John R Pierce -
John R. Dennison -
Jussi Hirvi -
Jussi Hirvi -
Keith Roberts -
Lamar Owen -
Les Mikesell -
Ljubomir Ljubojevic -
m.roth＠5-cent.us -
Marko Vojinovic -
Nicolas Thierry-Mieg -
Rajagopal Swaminathan -
Robert Heller -
Robert Nichols -
Rudi Ahlers -
Ryan Wagoner -
Steven Crothers -
Thomas Harold -
Timothy Murphy -
夜神 岩男