For interest: Bugzilla performance

[Bradley Baetz]

On Wed, 6 Nov 2002, Myk Melez wrote:

> Note also that emails sometimes mention multiple killed queries, 
> especially when the queries are generated by a DOS attack.
> 
> We stopped a DOS attack last Wednesday evening, not the first we've 
> dealt with in the last month, and that may have skewn results both 
> before and afterwards.

Potential project for interested people - hack mod_throttle to allow
whitelisting ip blocks ;) (Does mod_throttle catch anyone anymore, since
the limit was raised to deal with NSCP's NATed connections?)

> Unfortunately we'll inherit some performance regressions as well from 
> some recent code changes (the groups rewrite in particular) that the 
> performance improvements will hopefully counteract.

Yeah. Unfortunately, there was no working arround that one - a bitwise AND
is always going to be quicker than an sql table join, no matter how you
write the query. It has given us some nice features, although Joel's stuff
isn't landing before bmo's upgrade, so most of the advantages are only
infrastructure related.

(The timing you did yesterday was on the whole script, mind you, not just
on the sql, so its possible, although unlikely, that something else slowed
down. We should test that at some point)

Bring on mod_perl! ;)

> -myk

Bradley