Re: OT - Blog entry on hugepages

From: kyle Hailey <kylelf@xxxxxxxxx>
To: Brandon.Allen@xxxxxxxxxxx
Date: Wed, 4 Apr 2012 13:08:23 -0700
than paging and scanning stats that might be solved with largepages?
  Recently an AIX customer  was seeing slow I/Os from an NFS filer but fast
local I/O response times.  Of course the customer blamed the NFS filer, but
the filer reported constant speeds during good periods as well as bad. I
suspected the network, but they were on 10GbE which looked to perform
well.  The customer claimed that on the database host box there was free
CPU and that there was no scanning or paging out during the periods of NFS
I/O slowdown.  I personally didn't see the scanning or paging data. They
were using some AIX tools that I wasn't familiar with and I didn't see
vmstat output. Once the went to large pages the NFS read I/O slowdown went
away, and they are convinced that it's the large pages that "fixed" the
problem.  Seems like it did, but I'm uncomfortable with the analysis. The
only indication of memory pressure problems the customer gave was high SYS
cpu which seems like a less than satisfactory way to identify a problem of
memory pressure.
   If I remember correctly, at "Brushco" (maybe some of the original
players can chime in here) the first time I ran into largepage issues was
on a Sequent almost 20 years ago where each process had a copy of the
addressable process memory page table, so the larger the SGA, the more room
these pagetables took up and the more databases process there were, the
more memory was consumed by these process memory tables .By going to large
pages the pagetable was smaller and less memory was used. This could be
seen with memory issues that showed up in scanning and page outs.

- Kyle
http://dboptimizer.com

On Thu, Mar 15, 2012 at 9:41 AM, Allen, Brandon
<Brandon.Allen@xxxxxxxxxxx>wrote:

> Just a small correction - the 64k pages in AIX are called "medium pages".
>  The "large pages" in AIX are 16M and they require some extra
> configuration.  For more detail:
>
> http://public.dhe.ibm.com/partnerworld/pub/whitepaper/162b6.pdf
>
> Regards,
> Brandon
>
> -----Original Message-----
> From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
> On Behalf Of Rich Jesse
>
> <snip>
>  AIX, where the 64k largepages are self-managed in both AIX and Oracle.
> <snip>
>
>
>
> ________________________________
>
> Privileged/Confidential Information may be contained in this message or
> attachments hereto. Please advise immediately if you or your employer do
> not consent to Internet email for messages of this kind. Opinions,
> conclusions and other information in this message that do not relate to the
> official business of this company shall be understood as neither given nor
> endorsed by it.
> --
> //www.freelists.org/webpage/oracle-l
>
>
>


--
//www.freelists.org/webpage/oracle-l
Follow-Ups:
- Re: OT - Blog entry on hugepages
  - From: Rich Jesse
Re: OT - Blog entry on hugepages

Other related posts:
