Here are some screenshots of the CPU and Wait patterns of a few databases. I will make some seperate pages where I will dig a more deeper in the graphs, and do a little analysis. These graphs is just to get an idea what data Phat can get out of your PERFSTAT schema, and it will give you an idea of some different characters of database usage.
Note: I used to divide time between CPU and IO time. But later on I though ‘wait’ would be a better name instead of ‘IO’, so I started the renaming. But I still have a lot of Excel graphs where the name IO is used.
Databases serving content for large internet sites.
Internet database 1
Day/night pattern is visible, also the weekends can be recognized as lower peaks.
Internet database 2
An interesting one. The system has been taken in production recently. One high peak might be a potential problem, more investigation needed. The other daily returning peaks are exports taken every day. Apart from that no day / night pattern can be seen, this is quite strange. Maybe this is because the site is internationally oriented, more investigation is needed.
Internet database 3
This graph is showing trouble. Lots of blocking locks on interfaces in the last part of the graph. Apart from that, daily returning peak must be an interface or a backup.
Databases for OLTP systems, used during office hours.
OLTP database 1
Clear day / night week pattern can be seen, and a few suspicious peaks.
OLTP database 2
Clear day / night / week pattern can be seen. Something suspicious occurs now and then.
OLTP database 3
Graph show a bit of an irregular pattern for an OLTP system used in office hours. Week pattern can be spotted with a little effort.
OLTP database 4
Very clear day / night / week pattern with some very suspicious peaks can be seen.
Here’s an irregular pattern of an OLAP database.