We had over 500 application agents, about 300 epagent, and 16 IBM MQ agents in our CA APM 10.0 environment. The MQ agents were the most metrics out of the groups with our application agents producing from 2000 to 6000 metrics, our epagent producing around 120 metrics while our MQ agents had a spread of 5000 to 39,000 metrics each. We had about 900k metrics on seven collectors, postgres database, all on virtual servers running SuSE.
The 10.0 environment is slowly being phased out with a 10.5.2 on RHEL and hopefully by the end of the month, we will decommission our 10.0 environment.
The big things that helped was to insure that our servers had direct read/write to the disk, DASD instead of having the OS cache both. This helped with the harvest and smartstor duration.
We had over 150k metrics on each of the collectors with a target goal of no more than 180 to 200k metrics so that if we lose two of our collectors, the five remaining could keep up.
Hope this helps,
Billy