VBR-Info: md=nordu.net; mc=all; mv=swamid.se
Date: Thu, 27 Aug 2015 11:35:31 +0200 (CEST)
From: Henrik Thostrup Jensen <htj@nordu.net>
To: Jeff Janes <jeff.janes@gmail.com>, Emre Hasegeli <emre@hasegeli.com>,
 pgsql-performance <pgsql-performance@postgresql.org>
Subject: Re: Gist indexing performance with cidr types
In-Reply-To: 
 <CAMkU=1x=puaY9M-xk6=9TZQbWe6EBSMDQFVGNPHGRrM9f_Gc7A@mail.gmail.com>
Message-ID: <alpine.DEB.2.11.1508271121170.3739@pyrite>
References: <alpine.DEB.2.11.1508251504160.31004@pyrite>
 <CAE2gYzyv8zCeKUKyE8EcZcVtdTFeZaUucRjRqmtzu2jT5EQQyQ@mail.gmail.com>
 <alpine.DEB.2.11.1508260940060.35330@pyrite>
 <alpine.DEB.2.11.1508261031400.35330@pyrite>
 <CAE2gYzxh0SVZe=nOsaLqfnNWKAi_HvDn4s+-J-co9aYwcw71Vw@mail.gmail.com>
 <alpine.DEB.2.11.1508261245250.35330@pyrite>
 <CAMkU=1x=puaY9M-xk6=9TZQbWe6EBSMDQFVGNPHGRrM9f_Gc7A@mail.gmail.com>
User-Agent: Alpine 2.11 (DEB 23 2013-08-11)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed
Received-SPF: neutral (e-mailfilter02.sunet.se: 109.105.110.42 is neither
 permitted nor denied by domain htj@nordu.net)
 receiver=e-mailfilter02.sunet.se; client-ip=109.105.110.42;
 envelope-from=<htj@nordu.net>; helo=smtp1.nordu.net;
 identity=mailfrom
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

On Wed, 26 Aug 2015, Jeff Janes wrote:

> Any chance you can share the actual underlying data?

Sure. I added a snapshot to the repo:
https://github.com/job/irrexplorer/blob/master/data/irrexplorer_dump.sql.gz?raw=true

> I noticed it wasn't on github, but is that because it is proprietary, or
> just because you don't think it is interesting?

I hoped it wouldn't be this complicated :-).

BGP and IRR data is (mostly) public, but it changes constantly, so there 
is little sense in putting in the repo, as it is not the authorative 
source (we have a script to boostrap with instead).


> If you loop over the 732 rows yourself, issuing the simple query against each retrieved constant value:
> 
> explain (analyze,buffers) select routes.route from routes where route && $1
> 
> Does each one take about the same amount of time, or are there some outlier values which take much more time than the others?

I wrote a small script to try this out. It queries for each route 20 times 
to try and suppress the worst noise. I've sorted the results by time and 
put it here: https://gist.github.com/htj/1817883f92a9cb17a4f8
(ignore the ntp timing issue causing a negative value)

Some observations:

- v6 is faster than v4 which is expected.

- The slowest prefixes by all seem to start bits '11'.
   However it is only by a factor of 1.5x which is not really significant


     Best regards, Henrik

  Henrik Thostrup Jensen <htj at nordu.net>
  Software Developer, NORDUnet


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance