My system (instance of ONLINE) and the machine (NCR 3435) are running very
slowly.
History:
on 4/23/98 we dumped the data from two of our major tables ("case" and
"person")
modified the text files to accomodate the change to the tables
rebuilt the tables with the modified schemas and reloaded data.
Then we rebuilt indexes.
The "case" table is in its own dbspace, the "person" table is in the dbspace
set aside for the rest of the database. The root dbspace is seperate from
the other two dbspaces. The only thing that I know of that is different from
before (other than two new fields) is that when I built the "case" table in
its own dbspce, I created the first extent with a size just a little smaller
than the size of the dbsapce at the suggestion of Informix tech support.
That seems to make sense to me: one big extent.
Another note: the "case" dbspace and the "production" dbspace (the rest of
the database) are on seperate slices, but on the same disk device, however
they were that way before the migration of the data took place.
Now, in addition to the users complaining about the speed of simsply
queries, if I run
the UNIX command "ps -ef|grep oninit" I see that the oninits associated with
this instance that are "cpu" threads have very long run times.
If I run the UNIX command "sar -d 1 10" to look at the devices, I see that
the device that the two dbspaces is on has very high average usage
(90 -100%). Obviously that indicates a slow down.
Why would the changes to those two tables that I made be slowing things
down? The new fields didn't increase the size of the data substantially.
Here is another piece of information: users are reporting that reports
(which almost always involve the "case" table) are being returned quickly.
That makes sense consdiering that the "case" data is contiguous. Why is the
device taking such a hit? (I am planning to move data around at some future
date, but it won't be easy due to limited resources)
Please respond by e-mail and post.
Thank you
Paul Finkel
AT&T
732-457-5170