Re: Performance Benchmarking for data-warehousing instance?

From: Tomas Vondra <tv(at)fuzzy(dot)cz>
To: pgsql-performance(at)postgresql(dot)org
Subject: Re: Performance Benchmarking for data-warehousing instance?
Date: 2014-02-09 15:42:37
Message-ID: 52F7A1ED.1050701@fuzzy.cz
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Hi,

On 8.2.2014 07:41, Huy Nguyen wrote:
> Do you happen to have a link to it? Though I think different machine
> specs should yield different optimal postgresql.conf.

An optimal configuration is not just about machine specs, it's about the
workload and application configuration too. So there's no benchmark that
would give you the best config for your application.

> I'm looking for a hand-crafted set of data + queries tailored for OLAP
> so that I can try to manually tweak one config at a time and run against
> the benchmark.

I think using pgtune is the best starting point you can get, and you may
tweak it based on your actual workload. If you can prepare a sample of
the workload (i.e. a representative amount of data) and run a set of
actual queries (generated by the application), that'd be an excellent
situation.

> I might considering creating one if no one has done it before.

So how exactly is that going to work? There's an benchmark for this,
called TPC-H [1], but again - this is just a model of how a DWH/DSS
application may look like.

I've spent a lot of time working with it a while ago (see [2]), and IMHO
the values recommended by pgtune are quite fine.

[1] http://www.tpc.org/tpch/default.asp
[2] http://www.fuzzy.cz/en/articles/dss-tpc-h-benchmark-with-postgresql/

regards
Tomas

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Tomas Vondra 2014-02-09 15:50:04 Re: Bloated tables and why is vacuum full the only option
Previous Message Wattana Hinchaisri 2014-02-08 08:24:59 Re: Performance Benchmarking for data-warehousing instance?