[CentOS] Centos/Linux Disk Caching, might be OT in some ways

Christopher Chan christopher.chan at bradbury.edu.hk
Fri Jan 29 02:13:44 UTC 2010


On Friday, January 29, 2010 08:50 AM, Ross Walker wrote:
> On Jan 28, 2010, at 7:27 PM, Christopher Chan<christopher.chan at bradbury.edu.hk
>   >  wrote:
>
>> On Thursday, January 28, 2010 10:48 PM, Ross Walker wrote:
>>>
>>> On Jan 27, 2010, at 7:50 PM, Christopher Chan<christopher.chan at bradbury.edu.hk
>>>> wrote:
>>>
>>>>
>>>>> Sorry to be the bearer of bad news, but on top of LVM on CentOS/
>>>>> RHEL
>>>>> the best assurance your going to get is fsync(), meaning the data
>>>>> is
>>>>> out of the kernel, but probably still on disk write cache. Make
>>>>> sure
>>>>> you have a good UPS setup, so the disks can flush after main power
>>>>> loss.
>>>>
>>>> Or turn off write caching...
>>>
>>> Have you tried doing any kind of write with write caching turned off?
>>> It is so horribly slow to make it almost useless.
>>
>> If they needed the performance in the first place, I doubt they
>> would be
>> using md raid1. You want performance and reliability? Hardware raid +
>> bbu cache. Otherwise, it is turn off write caching unless the i/o path
>> supports barriers.
>
> Yes, but a lot of people jumped on the SW RAID is just as good or
> better then HW RAID bandwagon and well there is no battery backed up
> write cache then.
>

This was so long long ago. Performance-wise, software raid was better 
than hardware, especially when it comes raid5, when them hardware raid 
cards were a joke using Intel i960 processors or had little to no cache 
like the 3ware 75xx/85xx cards. But even then, hardware raid was good 
for raid1 because at that time, on 3ware anyway, md raid1 did not 
support partial syncs and had to do full syncs if a disk fell off the 
array but hardware raid implementations did support and had less of a 
penalty.

You will only find software raid faster than hardware raid for small 
setups, 4 to 6 disks, with raid1+0 today before bus contention gets in 
the way unless you have a motherboard like the one the Thumper uses. 
Until RHEL6 comes out though, that would be moot for data integrity 
unless you have a nice big umem nvram card and turn write caching off if 
you use lvm too.


>
>>>
>>> If you need to turn write-caching off then I would start looking at
>>> SSD drives with capacitor based caches.
>>>
>>
>> How do those compare with bbu nvram cards for external data + metadata
>> journaling?
>
> Slightly slower then nvram, but don't suffer from the write-cache
> filling up under load.
>


Dunno about the nvram card filling up...depends on the usage. Ext3/4 
filesystems being used as mail queues will be laughing with an external 
nvram device for its journal in data=journal mode unless most mails 
actually have to queue or are not sent off within five seconds (which is 
still adjustable i believe).


More information about the CentOS mailing list