Database peformance after switching to CIO
Date: Thu, 24 Apr 2008 16:18:21 -0700 (PDT)
OS - AIX 3
Oracle - 188.8.131.52
We recently switched over from async IO to CIO. [filesystemio_options
Reason for this move was a lot of log buffer and log sync waits. Contrary to the expectation, we have noticed performance degradation in our batch jobs.
System Event Total Waits Timeouts Time Waited db file sequential read 234,932,597 0 516,917 db file scattered read 104,119,285 0 218,360 rdbms ipc message 38,010,738 1,642,294 5,860,312 PX Deq: Execution Msg 31,088,553 105,763 237,518 PX Deq: Execute Reply 24,608,715 2,489 84,274 direct path read 22,753,359 0 92,395 PX Deq: Table Q Normal 19,672,286 16,399 66,533 buffer busy waits 19,514,194 5,524 55,345
Though log buffer and log sync waits have disappeared, but sequential read and scattered read waits have increased. One recommendation that came forward was to increase db cache size. Reason behind this was as there is no OS level caching with CIO and oracle needs to manage it, we need a large db cache size.
I just wanted to know if anyone has run into a similar situation and if increasing db cache size is the right direction.
Atul Received on Thu Apr 24 2008 - 18:18:21 CDT