MySQL Forums
Forum List  »  NDB clusters

Re: Example of Cluster SLOWER than InnoDB -- help?
Posted by: John Muehlhausen
Date: December 07, 2004 02:41PM

I split the <queueName>_state table into three: one for get seqnums, one for put seqnums and one for the dirty flag. (Breaking the dirty update that depends on seqnums in the process, but I'm just messing around at this point.)

That really helped, Cluster is now 50% the speed of InnoDB. My guess is that the locking in NDB is much more complicated (and time consuming) in NDB since it involves message passing.

Profiling shows that a couple ms is spent waiting in each mysql_query() call. I guess the only way to get rid of that cumulative time is to execute the whole transaction with one mysqlclient API call -- which requires stored procedures?

Options: ReplyQuote




Sorry, you can't reply to this topic. It has been closed.

Content reproduced on this site is the property of the respective copyright holders. It is not reviewed in advance by Oracle and does not necessarily represent the opinion of Oracle or any other party.