Have a look at (and post here if needed) the trace of one such queries (https://www.monetdb.org/Documentation/Cookbooks/SQLrecipes/QueryTiming), so that you can see what the query actually does.

It sounds likely that your queries need to materialize to main memory large intermediate results before they complete.

This, together with possibly little main memory available on the server, can turn into disk I/O because of memory mapping.

On 30 January 2015 at 10:00, Tim Burress <tim@variosecure.net> wrote:

Hello,

I have a large database that is commonly used for relatively simple
SELECT queries. Some of these can take quite a bit of time (hours...
overnight) so I've been looking into it a bit. What seems clear is that
mserver5 (Oct2014-SP1) is disk-bound, but one curious thing is that most
of the time it is write-bound. Despite doing only SELECT queries (no
updates or inserts), the amount of data being written is nearly two
times as much as data read.

Why is that? Is the server restructuring the database on the fly in
order to get better performance in later queries? 3X the disk I/O seems
excessive so I wonder if there is some way to tune this, or investigate
further to see what is happening?

Thanks!

Tim

_______________________________________________
users-list mailing list
users-list@monetdb.org
https://www.monetdb.org/mailman/listinfo/users-list