Re: OLTP 10x improvement with exadata and flash write back cache
From: Saibabu Devabhaktuni <saibabu_d_at_yahoo.com>
Date: Thu, 25 Oct 2012 17:15:11 -0700 (PDT)
Message-ID: <1351210511.23512.YahooMailNeo_at_web161302.mail.bf1.yahoo.com>
Hrishy,
Assuming that all writes go to flash cache, most benefits could come from:
Date: Thu, 25 Oct 2012 17:15:11 -0700 (PDT)
Message-ID: <1351210511.23512.YahooMailNeo_at_web161302.mail.bf1.yahoo.com>
Hrishy,
Assuming that all writes go to flash cache, most benefits could come from:
- Out of row stored Lob (with no cache option) direct path writes.
- Direct path writes generated by sort's spilling to temp files.
- Direct path writes generated by Index build, bulk loads, CTAS, etc.
- Faster checkpoints and this would help physical standby catch up faster (if it was lagging) as log boundary checkpoints on the standby are done while media recovery wait on it. Oracle introduced couple of new parameters to influence log boundary checkpoint in 11.2.
Unless your buffer cache is not configured properly, foreground processes waiting on DBWR for free buffer waits is typically not seen and hence it may not be a big benefit.
Thanks,
Sai
http://sai-oracle.blogspot.com
-- http://www.freelists.org/webpage/oracle-lReceived on Fri Oct 26 2012 - 02:15:11 CEST