Re: analyzing, visualizing, understanding and rating I/O latency
Date: Thu, 2 Aug 2012 11:43:27 +0100
Message-ID: <CAGdek=wf_WTsemcX2ToMxQk42HcA2=ME12=b4KjeeDGMJj9o8g_at_mail.gmail.com>
On Mon, Jul 30, 2012 at 7:47 PM, kyle Hailey <kylelf_at_gmail.com> wrote:
> Two questions I'm interested in answering or getting opinions on are:
>
> 1. what is considered great, good, ok and bad I/O and why? what do you
> use? latency values, throughput values, spread (ie like stdev), percentile
> latency
Hi Kyle,
I'm generally looking into latency - for most of system disk queue size and queuing time is a good indicator of performance problems.
> 2. is average latency good enough and if not why not? how would one use
> a latency histogram or 95-99.99 percentile latency to judge an I/O
> subsystem?
Generally average is good enough for data files but I use both average
and histogram for redo log file systems.
In both cases histograms can be useful to detect storage problems
(like bad disk in array) - if you have millions of operations at 2 ms
level
and a few operations at 1 s level due to disk error (but number of
this operations is growing) it will take a while until average will be
impacted
but it is easy to catch in histograms - I found v$event_histogram very
useful in this area.
regards,
-- Marcin Przepiorowski http://oracleprof.blogspot.com -- http://www.freelists.org/webpage/oracle-lReceived on Thu Aug 02 2012 - 05:43:27 CDT