W3C home > Mailing lists > Public > www-rdf-interest@w3.org > July 2004

Re: SIMILE Store Report

From: Andrew Newman <andrew@tucanatech.com>
Date: Thu, 29 Jul 2004 21:00:48 +1000
Message-ID: <4108D8E0.4020802@tucanatech.com>
To: www-rdf-interest@w3.org

> First of all I'd like to thank Lee for the report, it's interesing
> reading.
> 

Yes I agree, although it would've been nice to have been involved in the 
process.  One of the reasons why we open sourced our code was to get 
feedback.  I would've like to have had a chance to fix problems with 
Kowari or the way the application was using Kowari.

For a simple thing, like loading the triples it's obvious that Kowari is 
faster (I'll leave out the local Jena model because it's not persistent):
* Jena w/ Postgres - 971784 ms
* Jena w/ MySQL 4 - 844257 ms
* Jena w/ MySQL 3 - 667138 ms
* Kowari - 139092ms
* 3Store - 213088ms

I'm not sure Kowari is going to be the best at everything, not at all, 
but it would be nice to find out where we suck.  As it is I can't be 
sure where the problems are.

> It could be that the queries are unusually small, in 3store the networking
> overhead can become the biggest factor if the queries are unusually simple,
> eg:
> 
> SELECT ?x WHERE (<a>, <b>, <c>), (<c>, <d>, ?x)
> 
> or things of that kind. The time to execute the query will be < 1ms, but
> the HTTP request overhead is close to 2ms.
> 

I agree, the queries performed can make a huge difference.  For example, 
if something like walk [1] was used to traverse up or down predicates 
(like subClassOf) it would give a vastly different result compared with 
doing it with individual queries.

[1] http://www.kowari.org/1532.htm
Received on Thursday, 29 July 2004 07:01:07 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:14:58 UTC