Re: [GNUnet-developers] MYSQL optimization

[Hendrik Pagenhardt]

On Monday 15 December 2003 17:47, Igor Wronsky wrote:
> Sounds interesting. On a quick look it seems to require
> 4.x mysql.

I don't see where it does use features which where not used in the 
original version? Hmmm.... I only have mysql 4.0.xx here, so I can't 
verify the correct function on 3.x.

> and your method really is faster, I see no reason for
> not adopting it.

Well on my setup (6GB quota, 1.2GB used, 512MB memory, Athlon 700) it 
reduces the query time from around 1-2 seconds to about 0.01 seconds. And 
it reduces the constant hard disk activity I had after I switch to mysql 
to the occasional flickering that I had with gdbm.

The main reason for the speedup is that my method does only an index match 
(which is very fast, as the indexes should be cached) while the other 
method collects on average 128 rows from the data table (depending on 
fill level of the buckets) and than assign random values to them, sorts 
them and then selects the first one. It's a real pity, that relational 
databases have no simple means of accessing random rows...

Maybe some people can try my version and report the effects on there 
systems.

Ciao,
        Hendrik