« Will Yahoo! be next to support Jabber/XMPP IM? | Main | Market Research the Easy Way (Fashionistas take note!) »

August 27, 2005

Comments

Martin Geddes

I ought to say that the grid stuff was tongue in cheek -- I wasn't making any specific architectural recommendations.

I'm amazed at how few web sites use the prospective method of page construction (disk space is real cheap!), and make you laboriously sit through some cruddy RDBMS whirring away.

Keep up the good work,
Martin

Matthew Hurst

One thing to keep in mind is the query that is used to carry out these performance tests. I have been told that the term 'new york' is used in all queries. I doubt that this is in any way representative of either the general type of query, or the general volume of results. Search engines are tuned to the most common type of query/results, not a single arbitrary query like this.

I'm not apologising for the perfomance of any of the systems profiled here, but I am asking that we be a little more critical of information that we find out there.

(I'm posting this via Anonymizer as, ironically, I can't seem to access Bob's blog from where I am just now.)

Stephen Pierzchala

Bob:

Thanks for the pickup on the GrabPERF Measurement System. I have just released the first GrabPERF Search Performance Index, which is available here:

http://www.newestindustry.org/index.php/2005/08/29/grabperf_search_index_weekly_results

I do agree with Matthew Hurst's comment above that the query for "new york" may not be repreentative. However, if you want to set a baseline for comparison, this is the place to start.

In my other life, I do comparisons on a much broader subset of performance types and have had a hand in creating the concepts behind Consistency and Optimization indices. (http://www.gomez.com/performance_strategies/benchmarking.html)

I'm always open to feedback and look forward to your continued support.

smp

The comments to this entry are closed.