On Wed, Mar 11, 2009 at 12:55 PM, Joseph L. Casale
<JCasale@xxxxxxxxxxxxxxxxx> wrote:
>>Life is much better now that I'm using -h instead of the manual.
>>
>>I have a new question about policies
>>
>>Direct and DisDskCache makes sense
>>Cached and EnDiskCache makes sense
>>
>>not so sure about
>>
>>Direct and EnDskCache
>>Cached and DisDsk Cache
>>
>>Do they make sense?
>
> I believe Cached|Direct refers to adapter based caching of the logical
> (virtual) disk whereas EnDskCache|DisDskCache refers to physical disk
> cache.
>
> The adapter can hold data in the event of a powerloss with a BBU
> but I can't see how the disc could.
>
Hmm. The drive cabinets could have their own UPS. Doesn't have to be very
big to keep the power up for long enough for the drive disk caches to
get emptied.
And I've seen data centers with amaizing multiple layers of redundant power.
I'm thinking production database Secondary servers should be
configured for reliability, and production Primary servers should be
configure for speed.
The secondary will be extremely paranoid. WT Direct DisDskCache.
Not sure understand the performance benefit/risk trade offs well
enough to choose intelligently for the primary. Especially when the
application may be making lots of assumptions about what is happening
on the lower levels. And may have much more memory available to do
its own caching and optimization.
Does the controller really know enough about what's going on in the
drive to make effective use of the drives cache to speed things up?
My guess is that I'll start with WB Cached DisDskCache on the primary,
and may be surprised where we end up.
And it seems to me that ADRA is almost always the better choice. But
I could easily be missing something important.
--
Drew Einhorn
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
