Re: OLTP 10x improvement with exadata and flash write back cache

From: Kevin Closson <ora_kclosson_at_yahoo.com>
Date: Thu, 25 Oct 2012 15:12:34 -0700 (PDT)
Message-ID: <1351203154.13275.YahooMailNeo_at_web162801.mail.bf1.yahoo.com>



The photo shows "10x more write I/O's"

To drive DBWR at a rate of 10x you need CPU. That is, if free buffer waits are holding your processor-utilization at about 10% and your app (and what your app does to the Oracle database concurrency code) scales you might get 10x... but then don't forget that your observed processor utilization on modern processors (most of which have SMT) includes thread execution so a server that is at about 70% is generally maxed out. 

In other words, it depends.



 From: hrishy <hrishys_at_yahoo.co.uk>
To: "oracle-l_at_freelists.org" <oracle-l_at_freelists.org> Sent: Wednesday, October 24, 2012 11:19 PM Subject: OLTP 10x improvement with exadata and flash write back cache  

Hi
Has anybody started using this feature https://twitter.com/fritshoogland/status/235146135084351488

of and experienced a 10X improvement in their OLTP applications by enabling flash write back cache ?

Under  what specific circumstances do you think that 10X improvement is possible. What bottleneck would motivate us to turn on the exadata flash write back cache feature

Can a SSD also achieve the same thing as a flash write back cache ?

--

http://www.freelists.org/webpage/oracle-l
--

http://www.freelists.org/webpage/oracle-l Received on Fri Oct 26 2012 - 00:12:34 CEST

Original text of this message