Thanks for the views and I will consider the compressed tables (I haven't
read up on it properly but I assume I can run an alter for this; with the
usual dump/restore required as I am not using 1 file per table mode).

However in retrospect I wish I'd started with the same version of mysql and
tested and then upgraded as running a selection of random user generated
queries are all much slower on the new server, for instance:

My production server is Xeon(R) X3440  @ 2.53GHz with 5gig pool; ssd
storage; freebsd 8.1; zfs
New server is a Xeon(R) E5-2665 0 @ 2.40GHz with 40gig pool; ssd storage;
freebsd 9.2; zfs

The simple query select  count(*) from dbmail_fromfield; takes 8seconds on
the new and 3 seconds on the old.

A scan query of select * from dbmail_fromfield where fromfield like
"%EXAMPLE%"; takes 9 seconds on the new and 3.3 seconds on the old.

I have a feeling a steep learning curve is ahead to figure out what is going
on, the only thing I can think of is that as the production server is
running select queries constantly it's pool may be more useful but given the
new server is currently running as a replication slave I would have thought
that effect was irrelevant.

What gives me a little hope is that sysbench running it's oltp test is much
faster on the new machine (10,000 queries in 44 seconds with 1 thread vs 90
seconds on old) and scales well with additional threads.

Also of course these are not queries which dbmail actually performs when it
is running for real, and I have read dbmail 5.1 is faster for single
threaded queries.

Daniel

Reindl Harald (mobile) Tue, 22 Oct 2013 09:03:33 -0700 
5.5 is clearly faster and stable over years in the meantime - I even forgot 
which year we switched but it was painless 



_______________________________________________
DBmail mailing list
[email protected]
http://mailman.fastxs.nl/cgi-bin/mailman/listinfo/dbmail

Reply via email to