| From: | Joe Conway <mail(at)joeconway(dot)com> |
|---|---|
| To: | Justin Clift <justin(at)postgresql(dot)org> |
| Cc: | Dennis Björklund <db(at)zigo(dot)dhs(dot)org>, Gavin Sherry <swm(at)linuxworld(dot)com(dot)au>, Christopher Kings-Lynne <chriskl(at)familyhealth(dot)com(dot)au>, Paul Ramsey <pramsey(at)refractions(dot)net>, pgsql-advocacy(at)postgresql(dot)org |
| Subject: | Re: Elocution |
| Date: | 2002-12-10 06:48:35 |
| Message-ID: | 3DF58E43.8040909@joeconway.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-advocacy |
Justin Clift wrote:
> The OSDB software (another Open Source database testing tool) has added the
> ability to generate large datasets in recent CVS, so that's another avenue
> for testing with as well.
Just one note on this. If we ever get serious about using OSDB to compare
Postgres with MySQL, it needs some hacking in order to get a fair comparison.
The problem as I recall is that the two longest running tests (and therefore
the ones with the largest impact on the overall result) are designed to fetch
rows one-at-a-time. The Postgres version does this by using a CURSOR and FETCH
1, while the MySQL version has to fake it using the MySQL client library to
buffer the rows. In other words, the MySQL version of this benchmark is uses a
local library and cache, instead of making a round trip to the server for each
row (something like 10000 times)!
To make the comparison fair, the Postgres version should FETCH the entire
result once using libpq, and then dole out the rows one-at-a-time similar to
the MySQL version.
Joe
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Justin Clift | 2002-12-10 06:56:06 | Re: Elocution |
| Previous Message | Justin Clift | 2002-12-10 06:17:35 | Re: Elocution |