Re: Update Statements and Exadata
From: Saibabu Devabhaktuni <saibabu_d_at_yahoo.com>
Date: Mon, 10 Dec 2012 01:22:15 -0800 (PST)
Message-ID: <1355131335.26680.YahooMailNeo_at_web161304.mail.bf1.yahoo.com>
Unlike parallel inserts with APPEND, parallel update require all the necessary blocks to be read into buffer cache for the update operation. As the select part of update statement is offloaded to storage cell for smart scan, what happens if some blocks require clean out. I came across this blog http://dbatrain.wordpress.com/2011/04/21/how-smart-is-your-smart-scan/ on block clean out.
Date: Mon, 10 Dec 2012 01:22:15 -0800 (PST)
Message-ID: <1355131335.26680.YahooMailNeo_at_web161304.mail.bf1.yahoo.com>
Unlike parallel inserts with APPEND, parallel update require all the necessary blocks to be read into buffer cache for the update operation. As the select part of update statement is offloaded to storage cell for smart scan, what happens if some blocks require clean out. I came across this blog http://dbatrain.wordpress.com/2011/04/21/how-smart-is-your-smart-scan/ on block clean out.
Does storage cell really do the clean out and write back the modified blocks or just do it in the memory like parallel queries typically do?
If storage cell is doing persistent clean out, does it do full clean out or only update the ITL?
Thanks,
Sai
http://sai-oracle.blogspot.com
-- http://www.freelists.org/webpage/oracle-lReceived on Mon Dec 10 2012 - 10:22:15 CET