Hi,
Better integration into the optimiser is work ongoing, in 4.1 and 5.0 the algorithms to
choose indexes are very static, however it should select an index before scanning always
but selecting which index is not very good, performance of joins in general is something
worked on improving currently.
Rgrds Mikael
Anton Aleksandrov wrote:
> Hello to all.
> I don’t know how to formulate subject, but here
> is our situation. We’ve got one big table with
> 2.5mln records. We also run a query which joins
> 5-6 tables. Initial query was hanging the whole
> box – mysqld process was trying to take all
> memory and all swap of the server. Join is through
> primary keys.
>
> We started to investigate and found out that
> “SHOW TABLE STATUSâ? always returns 100.
>
> And with that queryâ| The only way we managed to
> run it faster was to force it to use indexes. This
> really looks like cluster sometimes fails to
> select correct index to use. Is it worth to open a
> bug? May be someone could provide more detailed
> info on this?
>
> Thanks,
> Anton.
Mikael Ronstrom
Senior Software Architect, MySQL AB
My blog:
http://mikaelronstrom.blogspot.com