Re: 11gR2 smart flash cache

From: Jeremy Schneider <jeremy.schneider_at_ardentperf.com>
Date: Wed, 12 Jan 2011 16:35:57 -0600
Message-ID: <4D2E2CCD.2040002_at_ardentperf.com>



One quick note on this... direct reads do not always indicate temp (or even parallel). In 11g Oracle will sometimes start using direct path reads for serial full tablescans.

I recently observed this happening a lot with one of my clients. FYI, I believe that this particular database had an underconfigured SGA... might be related. Interestingly, because Oracle was doing so many FTS with direct path, the BCHR looked deceptively healthy.

-Jeremy

Steve Harville wrote:
> I have not tried this setting but I do have some experience with
> Oracle on flash drives.
> We are an EMC shop so most I/O is already cached (all writes and all
> sequential reads). The system cannot cache random reads so that is
> where I use flash drives. The temp tablespace can benefit the most
> from flash drives since it exhibits this read pattern. If "direct
> reads" are a large part of your total wait time then you can probably
> benefit from moving temp to flash.
>
> Steve Harville
>
> http://www.linkedin.com/in/steveharville
>
>
>
> On Wed, Jan 12, 2011 at 5:24 AM, Zhu,Chao <zhuchao_at_gmail.com> wrote:
>
>> hi, List friends,
>> Oracle has been promoting this flash cache as second tier cache for
>> oracle for a while; Just wondering whether anyone has used this within the
>> industry?
>> We typically have 30gb-60gb SGA supporting 1TB-4TB database; We found
>> usually 30gb or 60gb cache size does not really matter much for majority of
>> our database(some has big diff though, depends on workload profile/active
>> dataset); But a 300gb flashdisk might make huge difference, and help reduce
>> the IOPS load on the SAN side?
>>
>> Looking forward to industry experience;
>>
>> Thx
>>
>>
>>
>> --
>> Regards
>> Zhu Chao
>>

-- 
http://www.ardentperf.com
+1 312-725-9249

Jeremy Schneider
Chicago

--
http://www.freelists.org/webpage/oracle-l
Received on Wed Jan 12 2011 - 16:35:57 CST

Original text of this message