Comments on: Hit ratios (2)

By: Faulty Quotes 4 – Buffer Cache Hit Ratio (BCHR) « Charles Hooper's Oracle Notes

Tue, 22 Dec 2009 14:47:32 +0000

[...] http://jonathanlewis.wordpress.com/2007/09/05/hit-ratios-2/ (a very detailed blog entry) “Ratios are highly suspect for monitoring purposes. If you think a ratio is helpful, think carefully about whether you should be monitoring the two underlying values instead. [...]

By: Advertisement: Nominet are recruiting « jarneil

Advertisement: Nominet are recruiting « jarneil — Tue, 15 Jan 2008 09:05:38 +0000

[...] have also been surprised when an interview candidate mentioned the buffer cache hit ratio in response to a question regarding a good way of investigating users complaining about performance [...]

By: Debunking the Buffer Cache Hit Ratio « die Seilerwerks

Debunking the Buffer Cache Hit Ratio « die Seilerwerks — Wed, 31 Oct 2007 20:02:59 +0000

[...] Debunking the Buffer Cache Hit Ratio Jonathan Lewis: 1 Buffer Cache Hit Ratio: 0 [...]

By: Ari

Ari — Wed, 10 Oct 2007 03:11:43 +0000

Despite the BCHR being of limited value, I still find it useful to build a picture of the workload of the databases I monitor.
Recording the database BCHR over time has alerted me when something has changed, eg a new job that runs frequently without the required indexes. Sure, there are other statistics that could also have been used to detect changes, and the database BCHR doesn’t always alert me to a problem. (Response time / wait event pattern changes are my main method).
But… to extend the meaning of BCHR a little:
I use a BCHR at the session level to separate OLTP users from batch jobs.
I also use BCHR at the SQL and segment levels to provide a hint as to whether full scans or single block gets are used, or how well objects are cached, etc. For example, a small table that has a large percentage of the database’s of physical reads and a low BCHR could be a candidate for the keep buffer pool. (Assuming I can’t change the code or add an index, which is often the case with off-the-shelf applications).
I do use a lot of other stats too, but the ratio between logical and physical I/Os at various levels help me to understand what is happening in a database.

By: Jonathan Lewis

Jonathan Lewis — Sat, 22 Sep 2007 13:40:15 +0000

The newsgroup's defender of the BCHR seems to believe that the list of arguments I presented above was randomly selected with ulterior motives, rather than being a careful summary of the points he made in his posts. However, when I asked him to supply a copy of any parts of his argument that he felt I had unfairly omitted he declined to do so. So, dear reader, please feel free to examine every post made on this thread by Bob Jones, and let me know if you can find any part of his argument that I have failed to address.

By: Hit Ratios (3) « Oracle Scratchpad

Hit Ratios (3) « Oracle Scratchpad — Thu, 20 Sep 2007 11:56:52 +0000

[...] Statspack, Troubleshooting — Jonathan Lewis @ 11:56 am UTC Sep 20,2007 A little while ago I wrote some notes on why monitoring the buffer cache hit ratio (BCHR) was a pointless exercise. I’ve recently [...]

By: Objects to put in keep buffer « Oracle Logbook

Objects to put in keep buffer « Oracle Logbook — Tue, 18 Sep 2007 19:41:57 +0000

[...] For more information about buffer pool : http://jonathanlewis.wordpress.com/2007/09/05/hit-ratios-2/ [...]

By: Good Buffer Distribution « Oracle Logbook

Good Buffer Distribution « Oracle Logbook — Fri, 14 Sep 2007 15:30:18 +0000

[...] For more information about buffer pool : http://jonathanlewis.wordpress.com/2007/09/05/hit-ratios-2/ [...]

By: Mirjana

Mirjana — Wed, 12 Sep 2007 08:25:15 +0000

Excellent!

By: SeánMacGC

SeánMacGC — Tue, 11 Sep 2007 13:54:28 +0000

Highly instructive Jonathan, thanks.
One day we all might see the curvature of the horizon ;o)