In the last episode (May 10), Eric J. Schwertfeger said:
> Also, my initial testing shows that INNOBASE tables are even faster
> than MyISAM, at least for insertions, though the same tupples in a
> similar table seem to take up more room. Aside from the restrictions
> mentioned in 8.7.4 of the reference manual, are there any other
> gotcha's that I should be looking out for?
Here are two more minuses:
No ANALYZE TABLE - correct key and subkey cardinality counts work
wonders for complex queries (probably easy to add).
No OPTIMIZE TABLE - Any table that you constantly insert and delete
on is going to fragment and have data blocks with holes. Since all
Innobase data goes into a single global tablespace, there's currently
no way to defrag the space as a while besides dropping and recreating
Suggested improvements would be the addition of the COALESCE TABLESPACE
and REBUILD INDEX commands, and a way to say "this tablespace can only
be used for this table/index/database" (rebuild index is easy to
implement; table coalescing is harder).
> So far, MySQL is doing quite well at matching or exceding PostreSQL's
> performance, though oddly enough, MySQL is taking more memory to do its
> job. Not complaining, mind you.
I have found that MySQL stays within the memory limits you give it in
my.cnf; you might want to lower some of the buffer settings depending
on your workload. http://www.mysql.com/doc/S/e/Server_parameters.html
is a good starting point.