Re: Partitioning / Clustering - Mailing list pgsql-performance

From Josh Berkus
Subject Re: Partitioning / Clustering
Date
Msg-id 200505121033.41254.josh@agliodbs.com
Whole thread Raw
In response to Re: Partitioning / Clustering  (John A Meinel <john@arbash-meinel.com>)
Responses Re: Partitioning / Clustering
List pgsql-performance
People,

> In general I think your point is valid. Just remember that it probably
> also matters how you count page views. Because technically images are a
> separate page (and this thread did discuss serving up images). So if
> there are 20 graphics on a specific page, that is 20 server hits just
> for that one page.

Also, there's bots and screen-scrapers and RSS, web e-mails, and web services
and many other things which create hits but are not "people".  I'm currently
working on clickstream for a site which is nowhere in the top 100, and is
getting 3 million real hits a day ... and we know for a fact that at least
1/4 of that is bots.

Regardless, the strategy you should be employing for a high traffic site is
that if your users hit the database for anything other than direct
interaction (like filling out a webform) then you're lost.    Use memcached,
squid, lighttpd caching, ASP.NET caching, pools, etc.   Keep the load off the
database except for the stuff that only the database can do.

--
Josh Berkus
Aglio Database Solutions
San Francisco

pgsql-performance by date:

Previous
From: Josh Berkus
Date:
Subject: Re: tuning Postgres for large data import (using Copy from)
Next
From: Alex Stapleton
Date:
Subject: Re: Partitioning / Clustering