On Tue, Nov 08, 2005 at 08:04:54AM -0800, Bryan J. Smith wrote:
NAS offers safe concurent access (generally, there might be some NAS devices outthere that do not). NAS device will manage file system internally, and export it over NFS or SMB protocols to the clients.
Such NAS' are a combined host+storage aka "filer." They have many advantages over SAN -- especially in their fail-over and/or load-balancing capabilities.
...
Oh, it all depends on the design of the NAS. NetApp does a pretty damn fine job with their designs (long story). But there's many other benefits. But that is a larger discussion.
Hi Bryan,
Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
I'm doing CentOS on servers these days, but I presume ext3 is not the best choice in this case. Previous postings of yours suggest XFS is the way to go. However, it seems hard to find an enterprise class linux distro with XFS incorporated? And how does a FreeBSD solution compare to linux/ext3 or linux/xfs?
What are the considerations in case of a NAS filer instead of a raid-box connected to hostmachine?
Regards,
On Wed, 2005-11-09 at 11:00 +0100, Henk van Lingen wrote:
Hi Bryan, Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
NetApp is very costly per $ versus traditional file storage. But the Data OnTap OS with WAFL filesystem was basically designed by 2 of Sun's original NFS designers. WAFL works very different than most traditional UNIX server filesystems.
The WAFL filesystem has a couple of different modes for network filesystem protocol access. One catered towards NFS, another catered towards SMB -- but you can access from both simultaneously, there are just considerations.
Solaris/x86-64 has a good bang-for-the-buck, and UFS supports quotas as well as Samba 3 ACLs. I haven't used their new filesystem with Samba though (anyone, anyone?).
I'm doing CentOS on servers these days, but I presume ext3 is not the best choice in this case. Previous postings of yours suggest XFS is the way to go. However, it seems hard to find an enterprise class linux distro with XFS incorporated?
Unfortunately, I'm finding it difficult to recommend XFS on Linux at this point. Not until Red Hat gets serious about it.
And how does a FreeBSD solution compare to linux/ext3 or linux/xfs?
XFS is being ported to FreeBSD, as some of the licensing issues have been worked out. But I wouldn't trust it anywhere close to even Linux at this point.
I'm a little outta date on FreeBSD and Samba, the last time I used Samba on FreeBSD was version 2.2 several years ago (yes, yes, I know, quite hypocritical for the guy who wrote the BSD appendix in "Samba Unleashed" -- but that was 5 years ago).
What are the considerations in case of a NAS filer instead of a raid-box connected to hostmachine?
Nothing really. I mean, you traditionally don't have full shell/filesystem access in a NetApp filer, and you need a "sister admin system" (one system with special mounts) to administer some "/etc" files, but otherwise, they are pretty nice.
Especially for fail-over, but it'll cost you.
On Wed, 9 Nov 2005, Bryan J. Smith wrote:
On Wed, 2005-11-09 at 11:00 +0100, Henk van Lingen wrote:
Hi Bryan, Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
NetApp is very costly per $ versus traditional file storage. But the Data OnTap OS with WAFL filesystem was basically designed by 2 of Sun's original NFS designers. WAFL works very different than most traditional UNIX server filesystems.
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
On 11/9/05, Robin Mordasiewicz robin@bullseye.tv wrote:
On Wed, 9 Nov 2005, Bryan J. Smith wrote:
On Wed, 2005-11-09 at 11:00 +0100, Henk van Lingen wrote:
Hi Bryan, Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
NetApp is very costly per $ versus traditional file storage. But the Data OnTap OS with WAFL filesystem was basically designed by 2 of Sun's original NFS designers. WAFL works very different than most traditional UNIX server filesystems.
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
Anyone considered or used OpenFiler? It's based on CentOS. http://www.openfiler.com/
-- Leonard Isham, CISSP Ostendo non ostento.
Leonard Isham leonard.isham@gmail.com wrote:
Anyone considered or used OpenFiler? It's based on CentOS. http://www.openfiler.com/
Damn that looks sweet! I didn't even know about it! CentOS 3-based.
I'm sure they haven't addressed failover and many other concerns, and it's not an "all-in-one" like old SME Server (fka E-Smith), but it's definitely a nice, entry-level "Filer".
Screenshots to note ...
Network/host definitions and access control: http://www.openfiler.com/screenshots/shots/general_local_networks_1.png http://www.openfiler.com/screenshots/shots/shares_edit_share_3.png
Select directory service(s): http://www.openfiler.com/screenshots/shots/accounts_authentication_2.png
Group and host access control: http://www.openfiler.com/screenshots/shots/shares_edit_share_5.png
Quotas: http://www.openfiler.com/screenshots/shots/quota_group_quota_2.png
I saw snapshots mentioned as well, which work quite well under kernel 2.4 and LVM (i.e., CentOS 3).
All screen shots: http://www.openfiler.com/screenshots/
FAQs -- General and Administration: http://www.openfiler.com/docs/faq/openfiler-general-faq.html
http://www.openfiler.com/docs/faq/openfiler-administration-faq.html
"Bryan J. Smith" thebs413@earthlink.net wrote:
Damn that looks sweet! I didn't even know about it! CentOS 3-based. I'm sure they haven't addressed failover and many other concerns, and it's not an "all-in-one" like old SME Server (fka E-Smith), but it's definitely a nice, entry-level "Filer".
Actually, they have quite a bit in progress: http://www.openfiler.com/docs/roadmap.html
Wow! I'm very impressed. Now I have a Linux solution I can recommend to SMBs when I don't have time (or they don't have the money for me ;-) to support them.
Damn that looks sweet! I didn't even know about it! CentOS 3-based. I'm sure they haven't addressed failover and many other concerns, and it's not an "all-in-one" like old SME Server (fka E-Smith), but it's definitely a nice, entry-level "Filer".
I think there is new life in the e-smith project. We are running A test on the latest here and seems to work quite well for A small office setup. Also joined their list, and updates/fixes Are coming out readily. See http://contribs.org
Gerald
Gerald Waugh gwaugh@frontstreetnetworks.com wrote:
I think there is new life in the e-smith project.
I never said there wasn't. ;->
And it's great to see CentOS as a staple of all these new "thin server" projects.
Bryan J. Smith wrote:
Leonard Isham leonard.isham@gmail.com wrote:
Anyone considered or used OpenFiler? It's based on CentOS. http://www.openfiler.com/
Damn that looks sweet! I didn't even know about it! CentOS 3-based.
I'm sure they haven't addressed failover and many other concerns, and it's not an "all-in-one" like old SME Server (fka E-Smith), but it's definitely a nice, entry-level "Filer".
Screenshots to note ...
Network/host definitions and access control: http://www.openfiler.com/screenshots/shots/general_local_networks_1.png http://www.openfiler.com/screenshots/shots/shares_edit_share_3.png
Select directory service(s): http://www.openfiler.com/screenshots/shots/accounts_authentication_2.png
Group and host access control: http://www.openfiler.com/screenshots/shots/shares_edit_share_5.png
Quotas: http://www.openfiler.com/screenshots/shots/quota_group_quota_2.png
I saw snapshots mentioned as well, which work quite well under kernel 2.4 and LVM (i.e., CentOS 3).
All screen shots: http://www.openfiler.com/screenshots/
FAQs -- General and Administration: http://www.openfiler.com/docs/faq/openfiler-general-faq.html
http://www.openfiler.com/docs/faq/openfiler-administration-faq.html
OpenFiler2 based on CentOS4/lvm2 is due out before the end of this month.
We are going to still support the CentOS3/Openfiler 1 series for a while, SnapShots being an important factor.
- K
PS: OpenFiler is my day job
Karanbir Singh wrote:
OpenFiler2 based on CentOS4/lvm2 is due out before the end of this month.
We are going to still support the CentOS3/Openfiler 1 series for a while, SnapShots being an important factor.
because this got asked offlist, right away..... SnapShots on Lvm2 are still kinda broken - the results you achieve are never predicatable.
During LinuxWorld London, I spoke to agk about this issue - his response was 'Snapshots are high on the agenda for lvm2, will work soon'.
- K
PS: OpenFiler is my day job
Actually, its not entirely. I work for Xinit Systems, and they in turn sponsor the Openfiler project. Offshoot being that I get to spend daytime office hours working on OpenFiler.
On 11/9/05, Karanbir Singh mail-lists@karan.org wrote:
because this got asked offlist, right away..... SnapShots on Lvm2 are still kinda broken - the results you achieve are never predicatable.
How broken? Any details? I found a few comments online that multiple snapshots may not work or that snapshotting can be memory intensive; will LVM2 snapshots work at all, or am I better off just avoiding them for now?
Josh Kelley
On Wed, 2005-11-09 at 14:09 -0500, Josh Kelley wrote:
On 11/9/05, Karanbir Singh mail-lists@karan.org wrote:
because this got asked offlist, right away..... SnapShots on Lvm2 are still kinda broken - the results you achieve are never predicatable.
How broken? Any details? I found a few comments online that multiple snapshots may not work or that snapshotting can be memory intensive; will LVM2 snapshots work at all, or am I better off just avoiding them for now?
I would not use snapshots at all on LVM2 right now (at least not for anything important, or on a machine that is important.
Josh Kelley joshkel@gmail.com wrote:
How broken? Any details?
Snapshots are not included in the current LVM2, unlike LVM.
You have to use more of the experimental DeviceMapper (DM) work going on. It will eventually be a very powerful component of kernel 2.6 -- much more than LVM on kernel 2.4.
But until then, avoid DM with snapshots. ;->
I found a few comments online that multiple snapshots may not work or that snapshotting can be memory intensive;
The big issue is that there are many "race conditions" in the VFS layers related to MD, LVM2, DM, etc... of kernel 2.6.
will LVM2 snapshots work at all, or am I better off just avoiding them for now?
Avoid.
When Red Hat says they are good, they will be included in their RHEL kernels. Until then, avoid. ;->
On 11/9/05, Bryan J. Smith thebs413@earthlink.net wrote:
Snapshots are not included in the current LVM2, unlike LVM.
When Red Hat says they are good, they will be included in their RHEL kernels. Until then, avoid. ;->
I'm confused?
Snapshots are enabled on my CentOS 4 boxes, and I've used them before in testing. Those are using LVM2 and (mostly) RHEL kernels. Unless I'm missing something...?
I appreciate the info, and I'll certainly avoid snapshots for now, I just didn't understand these comments.
Josh Kelley
Josh Kelley wrote:
Snapshots are not included in the current LVM2, unlike LVM.
When Red Hat says they are good, they will be included in their RHEL kernels. Until then, avoid. ;->
I'm confused?
Snapshots are enabled on my CentOS 4 boxes, and I've used them before in testing. Those are using LVM2 and (mostly) RHEL kernels. Unless I'm missing something...?
snapshots are available, but not supported - and the developers themselves say snapshots on lvm2 are broken.
eg. https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=164959 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=132057 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=168824
On 11/9/05, Karanbir Singh Mail-Lists@karan.org wrote:
snapshots are available, but not supported - and the developers themselves say snapshots on lvm2 are broken.
eg. https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=164959 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=132057 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=168824
Great, thanks for the links.
Josh Kelley
Leonard Isham wrote:
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
My EMC sales guy (who will be here in a bit) is trying to sell me a Celera NS500 series NAS to front-end a Clarion. I've worked with both NetApp and Celerra's in the past and prefer the separation in the EMC model from the disk-subsystem. Both are great for high-performance/high-availability CIFS/NFS services. Never used either for iSCSI although that will likely be used on this system as we move into production.
Dave www.hornfordassociates.com
On 11/9/05, Dave Hornford OSD@hornfordassociates.com wrote:
Leonard Isham wrote:
The text below did not originate with me. I don't want credit for something I did not write, and I don't have the background to be able to state the vailidity of it.
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
My EMC sales guy (who will be here in a bit) is trying to sell me a Celera NS500 series NAS to front-end a Clarion. I've worked with both NetApp and Celerra's in the past and prefer the separation in the EMC model from the disk-subsystem. Both are great for high-performance/high-availability CIFS/NFS services. Never used either for iSCSI although that will likely be used on this system as we move into production.
Dave www.hornfordassociates.com _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- Leonard Isham, CISSP Ostendo non ostento.
Leonard Isham wrote:
On 11/9/05, Robin Mordasiewicz robin@bullseye.tv wrote:
On Wed, 9 Nov 2005, Bryan J. Smith wrote:
On Wed, 2005-11-09 at 11:00 +0100, Henk van Lingen wrote:
Hi Bryan, Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
NetApp is very costly per $ versus traditional file storage. But the Data OnTap OS with WAFL filesystem was basically designed by 2 of Sun's original NFS designers. WAFL works very different than most traditional UNIX server filesystems.
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
Anyone considered or used OpenFiler? It's based on CentOS. http://www.openfiler.com/
I have had some contact with the company. Frankly, I thought they were rather expensive when you consider what's under the hood. I also wasn't too thrilled with their idea of service...."uh, if your system fails you can ship it back to us in the UK and we'll ship you another one." But maybe they've gotten better.
Cheers,
Chris Mauritz wrote:
Leonard Isham wrote:
On 11/9/05, Robin Mordasiewicz robin@bullseye.tv wrote:
On Wed, 9 Nov 2005, Bryan J. Smith wrote:
On Wed, 2005-11-09 at 11:00 +0100, Henk van Lingen wrote:
Hi Bryan, Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
NetApp is very costly per $ versus traditional file storage. But the Data OnTap OS with WAFL filesystem was basically designed by 2 of Sun's original NFS designers. WAFL works very different than most traditional UNIX server filesystems.
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
Anyone considered or used OpenFiler? It's based on CentOS. http://www.openfiler.com/
I have had some contact with the company. Frankly, I thought they were rather expensive when you consider what's under the hood. I also wasn't too thrilled with their idea of service...."uh, if your system fails you can ship it back to us in the UK and we'll ship you another one." But maybe they've gotten better.
you seem to be confused... openfiler is an open source distro built on CentOS. hardly the kind of thing you would need to 'ship to the UK'
- K
Karanbir Singh wrote:
I have had some contact with the company. Frankly, I thought they were rather expensive when you consider what's under the hood. I also wasn't too thrilled with their idea of service...."uh, if your system fails you can ship it back to us in the UK and we'll ship you another one." But maybe they've gotten better.
you seem to be confused... openfiler is an open source distro built on CentOS. hardly the kind of thing you would need to 'ship to the UK'
No, that was the result of an email exchange with Xinit which I found through the link section of the openfiler web site. Sorry about the confusion. I'll see if I can dig up the email. It came from a fellow with an Indian name (I think), but I can't remember it off the top of my head.
Cheers,
Chris Mauritz wrote:
I have had some contact with the company. Frankly, I thought they were rather expensive when you consider what's under the hood. I also wasn't too thrilled with their idea of service...."uh, if your system fails you can ship it back to us in the UK and we'll ship you another one." But maybe they've gotten better.
you seem to be confused... openfiler is an open source distro built on CentOS. hardly the kind of thing you would need to 'ship to the UK'
No, that was the result of an email exchange with Xinit which I found through the link section of the openfiler web site. Sorry about the confusion. I'll see if I can dig up the email. It came from a fellow with an Indian name (I think), but I can't remember it off the top of my head.
Someone told you to send the openfiler distro back to the UK for fixing if you had a problem ? I'd love to see that email :)
Anyway, OpenFiler is open source, you can download it and use it, the support issues are exactly the same as any other open source project.
Xinit provides other addon's ( like iscsi-target support ) and software support contracts, as a commercial service - using OpenFiler in no way forces the user to either buy or use this commercial service, but should you need it - its there.
- K
Karanbir Singh wrote:
Chris Mauritz wrote:
I have had some contact with the company. Frankly, I thought they were rather expensive when you consider what's under the hood. I also wasn't too thrilled with their idea of service...."uh, if your system fails you can ship it back to us in the UK and we'll ship you another one." But maybe they've gotten better.
you seem to be confused... openfiler is an open source distro built on CentOS. hardly the kind of thing you would need to 'ship to the UK'
No, that was the result of an email exchange with Xinit which I found through the link section of the openfiler web site. Sorry about the confusion. I'll see if I can dig up the email. It came from a fellow with an Indian name (I think), but I can't remember it off the top of my head.
Someone told you to send the openfiler distro back to the UK for fixing if you had a problem ? I'd love to see that email :)
Anyway, OpenFiler is open source, you can download it and use it, the support issues are exactly the same as any other open source project.
Xinit provides other addon's ( like iscsi-target support ) and software support contracts, as a commercial service - using OpenFiler in no way forces the user to either buy or use this commercial service, but should you need it - its there.
No. After reviewing the openfiler web site, I contacted Xinit and then emailed the sales contact about pricing and availability for an actual server. I've looked for the mail and can't find it. Perhaps one of your colleagues has a copy. It was probably on or about the summer of 2004. My problem with using the Xinit solution was two-fold. 1. I thought it was somewhat expensive compared to "rolling my own.". 2. When I inquired about service, I was told they didn't have anything available in North America and that I'd have to physically return the box to them and they'd send me a new one if I had a hardware failure. That didn't strike me as being terribly useful in an enterprise environment. My "maybe they've gotten better" comment was in hopes that they've since found a way to more quickly service clients in North America that experience hardware problems with an integrated (by Xinit) server.
I hope that clears up any confusion.
Cheers,
Robin Mordasiewicz robin@bullseye.tv wrote:
I will attest that NetApp is an excellent choice.
They *KNOW* NFS, that's for sure. ;-> 99.9% of other NAS devices don't (despite marketing).
NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them,
Not to nit-pick, you probably shouldn't (forgive me ;-). The media often likes to, but it's really not one to make.
In a nutshell, the EMC CLARiiON line spanning NAS-to-SAN fills a different, more flexible role. Sure, it can do many things NetApp filers can. But for most Small & Medium Business (SMB) "filer" needs, NetApp is far better for the task -- at a much lower price point.
When you deploy an EMC solution, you're typically solving more needs than just filers. So the more SMB "filer" capabilities of CLARiiON are basically just so you don't have to go out and buy a dedicated "filer" solution.
But yes, when you just want "filers," you want NetApp over EMC. I agree that you shouldn't let anyone tell you otherwise.
[ DISCLAIMER: I've worked both for and in conjuction with EMC at clients in the past. ]
and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone.
No one has come up with anything remotely capable to their Data OnTap OS with its WAFL filesystem. The overwhelming majority of NAS devices are typically based on BSD, Linux or Windows Server for Appliances, using their underlying services and filesystems.
I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform,
CLARiiON will, but at quite a price-point. But as I mentioned, if you're going CLARiiON, you're probably doing more things than just filers.
but there may be. I have used the RedHat Clustering solution and it was a hassle.
Compared to NetApp? No joke! You're not going to find anything PC/NOS based to be remotely as easy and scalable. In the best case, you might be looking to high-end Sun servers -- and then you should probably be considering EMC solutions if you need that scalability anyway.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
What's not to like (other than the price)?
Just so you know where it starts, if you're looking to go with NetApp filers for NFS and SMB, it's hard to start below $15,000, and plan at least $30,000 for an entry-level, redundant configuration.
Robin Mordasiewicz wrote:
On Wed, 9 Nov 2005, Bryan J. Smith wrote:
On Wed, 2005-11-09 at 11:00 +0100, Henk van Lingen wrote:
Hi Bryan, Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
NetApp is very costly per $ versus traditional file storage. But the Data OnTap OS with WAFL filesystem was basically designed by 2 of Sun's original NFS designers. WAFL works very different than most traditional UNIX server filesystems.
I will attest that NetApp is an excellent choice. NetApp is more pricey, but they do have an entry level system called the FAS270, which is leaps ahead of an EMC Clarion if you can even compare them, and for the added flexibilty the price is not very much more. I think NetApp pretty much stands alone. I don't think another solution will allow you to do CIFS, NFS, iSCSI, and Fibre Channel all from the same platform, but there may be. I have used the RedHat Clustering solution and it was a hassle.
P.s. I have nothing to do with the NetApp company, but I just love their product after having used it in almost every place I have worked at.
Netapp filers are wonderful. The ability to have a fast redundant file system that can periodically take snapshots of your data is a very good thing to have, especially in industries where you may be called upon to "roll back" your data to specific points in time on demand. They are also extraordinarily expensive. Same goes for EMC solutions. They are both doing very well on Wall Street applications these days and that's where I typically run into them. And in those environments, their cost is more or less statistical noise in the overall IT budget. For your typical small business or even mid-size business they're overkill and overpriced. For myself and on a few consulting gigs, NFS and samba have served me well, depending on the application, running on vanilla RHEL/CentOS systems with a pile of SATA disks hanging off 3Ware cards.
Cheers, .
Chris Mauritz chrism@imntv.com wrote:
Netapp filers are wonderful. The ability to have a fast redundant file system that can periodically take snapshots of your data is a very good thing to have,
That's the power of an OS and filesystem that is built with specific hardware features in mind -- such as NVRAM.
especially in industries where you may be called upon to roll back" your data to specific points in time on demand.
Er, yes and no. Yes, snapshots are great. But no, they don't replace full file revision control, databases, archiving, etc...
If you want full revision control built into the filesystem, you want to run VMS -- hence why it's still very popular. ;->
Hi Henk:
xfs is in the Centosplus repository. On the "enterprise class" if you mean "Redhat derived" then you have a point. However, SuSE and most of the other major distributions have native xfs support (and have had it for years). Redhat is rather alone in this regard, and this may be due to all their investment in ext3.
Joe
Henk van Lingen wrote:
Hi Bryan,
Please tell. I have to replace our old Sun Enterprise fileserver (solaris8), which does NFS and Samba (homedirectories, projects file space). It will be x86 hardware, but I'm looking for the best filesystem for the job (let's say one terabyte). It has to have quota and ACL support.
I'm doing CentOS on servers these days, but I presume ext3 is not the best choice in this case. Previous postings of yours suggest XFS is the way to go. However, it seems hard to find an enterprise class linux distro with XFS incorporated? And how does a FreeBSD solution compare to linux/ext3 or linux/xfs?
What are the considerations in case of a NAS filer instead of a raid-box connected to hostmachine?
Regards,
Joe Landman landman@scalableinformatics.com wrote:
xfs is in the Centosplus repository. On the "enterprise class" if you mean "Redhat derived" then you have a point. However, SuSE and most of the other major distributions have native xfs support (and have had it for years).
XFS working correctly an completely is another matter entirely. It was much better back in the days when SGI officially released XFS for select Red Hat Linux releases for kernel 2.4. The XFS support in the stock kernel has always been suspect (especially the 2.4 backport).
As someone who has been on the XFS lists over the last 5+ years, especially early on, SGI only supported XFS in its official releases (for Red Hat Linux). There was always massive breakage in various distros. SuSE is no exception (and don't get me started on Mandrake ;-).
E.g., SuSE has never been known for their attention to NFS compatibility. At one point in 2000, one SuSE engineer said I was much better off with Ext3 on Red Hat than ReiserFS on SuSE. Every now and then the Red Hat v. SuSE debate comes up on the XFS list and you'll quickly note people who have had nightmares with XFS on SuSE's distros.
Redhat is rather alone in this regard, and this may be due to all their investment in ext3.
Or the fact that Red Hat actually supports what it ships. SuSE has bit me in the @$$ too many times on NFS (let alone other distros). If you don't need NFS services, great! If you do, I would deter you from anything but Red Hat (or Sun ;-) in a distribution release.
At the same time, I agree that the lack of Red Hat interest in XFS is rather troubling. Especially the insistence that Ext3 can do everything XFS -- and those statements go silent when I start talking about everything from storing EAs in dumps to scalability to defragmentation. I documented that in my past blog entry here: http://thebs413.blogspot.com/2005/08/filesystem-fundamentals-and-practices.h...
But there are some real issues with XFS on 4K stack kernels and NFS compatibility right now. And I don't trust XFS in kernel 2.4, period (except for the older releases).
If you're going to run any XFS kernel, I recommend you pull a stable-tag'd version of the kernel out of XFS' CVS repository. How compatible/able it is with FC/RHEL/CentOS, I just don't know. But trying to add in patches for other things is not a nightmare I want to deal with. ;->
On Wed, 2005-11-09 at 06:43 -0800, Bryan J. Smith wrote:
Joe Landman landman@scalableinformatics.com wrote:
xfs is in the Centosplus repository. On the "enterprise class" if you mean "Redhat derived" then you have a point. However, SuSE and most of the other major distributions have native xfs support (and have had it for years).
XFS working correctly an completely is another matter entirely. It was much better back in the days when SGI officially released XFS for select Red Hat Linux releases for kernel 2.4. The XFS support in the stock kernel has always been suspect (especially the 2.4 backport).
As someone who has been on the XFS lists over the last 5+ years, especially early on, SGI only supported XFS in its official releases (for Red Hat Linux). There was always massive breakage in various distros. SuSE is no exception (and don't get me started on Mandrake ;-).
E.g., SuSE has never been known for their attention to NFS compatibility. At one point in 2000, one SuSE engineer said I was much better off with Ext3 on Red Hat than ReiserFS on SuSE. Every now and then the Red Hat v. SuSE debate comes up on the XFS list and you'll quickly note people who have had nightmares with XFS on SuSE's distros.
Redhat is rather alone in this regard, and this may be due to all their investment in ext3.
Or the fact that Red Hat actually supports what it ships. SuSE has bit me in the @$$ too many times on NFS (let alone other distros). If you don't need NFS services, great! If you do, I would deter you from anything but Red Hat (or Sun ;-) in a distribution release.
At the same time, I agree that the lack of Red Hat interest in XFS is rather troubling. Especially the insistence that Ext3 can do everything XFS -- and those statements go silent when I start talking about everything from storing EAs in dumps to scalability to defragmentation. I documented that in my past blog entry here: http://thebs413.blogspot.com/2005/08/filesystem-fundamentals-and-practices.h...
But there are some real issues with XFS on 4K stack kernels and NFS compatibility right now. And I don't trust XFS in kernel 2.4, period (except for the older releases).
The 4k stack problem are a major issue with XFS. I am working with someone at SGI to get some better code for our unsupported kernel, but that cods still has 4k stack issues.
I tried, very unsuccessfully, to get a RH patched 2.6.9-22.0.1 kernel to compile with 8K stacks.
To be honest, I would not use XFS in the CentOS kernel on a mission critical server.
The code we will roll in from SGI will be similar to the latest release on SuSE, but (as I said) that has 4k stack issues too.
I am sorry to say, ext3 is just the best and most stable bet.
The kernel in CentOS plus will run XFS, ReiserFS, and JFS ... but only ext2/ext3 are really rated as production stable (or the others would be in the standard kernel).
If you're going to run any XFS kernel, I recommend you pull a stable-tag'd version of the kernel out of XFS' CVS repository. How compatible/able it is with FC/RHEL/CentOS, I just don't know. But trying to add in patches for other things is not a nightmare I want to deal with. ;->
On Wed, 9 Nov 2005 at 9:53am, Johnny Hughes wrote
The 4k stack problem are a major issue with XFS. I am working with someone at SGI to get some better code for our unsupported kernel, but that cods still has 4k stack issues.
I tried, very unsuccessfully, to get a RH patched 2.6.9-22.0.1 kernel to compile with 8K stacks.
To be honest, I would not use XFS in the CentOS kernel on a mission critical server.
The code we will roll in from SGI will be similar to the latest release on SuSE, but (as I said) that has 4k stack issues too.
I am sorry to say, ext3 is just the best and most stable bet.
The kernel in CentOS plus will run XFS, ReiserFS, and JFS ... but only ext2/ext3 are really rated as production stable (or the others would be in the standard kernel).
Thanks a bunch for this update/summary. It says a lot about someone when they're willing to offer such an honest opinion of their "baby".
-
Bryan J. Smith -
Chris Mauritz -
Dave Hornford -
Gerald Waugh -
Henk van Lingen -
Joe Landman -
Johnny Hughes -
Josh Kelley -
Joshua Baker-LePain -
Karanbir Singh -
Leonard Isham -
Robin Mordasiewicz