Re: Initial prefetch performance testing - Mailing list pgsql-hackers

From Bruce Momjian
Subject Re: Initial prefetch performance testing
Date
Msg-id 200809241433.m8OEX6E29070@momjian.us
Whole thread Raw
In response to Re: Initial prefetch performance testing  (Gregory Stark <stark@enterprisedb.com>)
List pgsql-hackers
Gregory Stark wrote:
> 
> Ron Mayer <rm_pg@cheapcomplexdevices.com> writes:
> 
> > For example, on our sites hosted with Amazon's compute cloud (a great
> > place to host web sites), I know nothing about spindles, but know
> > about Amazon Elastic Block Store[2]'s and Instance Store's[1].   I
> > have some specs and are able to run benchmarks on them; but couldn't
> > guess how many spindles my X% of the N-disk device that corresponds
> > to.  
> 
> Well I don't see how you're going to guess how much prefetching is optimal for
> those environments either... 
> 
> > For another example, some of our salesguys with SSD drives
> > have 0 spindles on their demo machines.
> 
> Sounds to me like you're finding it pretty intuitive. Actually you would want
> "1" because it can handle one request at a time. Actually if you have a
> multipath array I imagine you would want to think of each interface as a
> spindle because that's the bottleneck and you'll want to keep all the
> interfaces busy.

I assume everyone would want at least one because you would want the
page to be prefetched while you are processing the existing page.  I
think a larger problem is that it is likely the page prefetch will take
longer than processing the existing page, which might mean that you need
to prefetch a few pages ahead to allow read reordering for better
performance.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + If your life is a hard drive, Christ can be your backup. +


pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: Initial prefetch performance testing
Next
From: Heikki Linnakangas
Date:
Subject: Re: Initial prefetch performance testing