We have Solaris servers with the CA SystemEdge agent monitored in Spectrum. We've run across the case with management agent lost conditions seemingly co-incident with a large number of running processes on the server. One server has at times greater than 5000 running processes. When attempting to view System Resources - Running and Monitored Processes - RFC 2790 - Running Processes table, it will take a long period of time to complete the queries and get responses needed to fill the table. Meanwhile, the SystemEdge agent will not respond to SNMP queries. I know that there may be limitations in resources that the server OS can provide to the agent. However, is this a design issue of the agent in that it's single threaded in execution of tasks?
This is likely an architectural limitation of the agent. It was designed to be light weight and is largely single-threaded. When dealing with higher numbers of processes in the RFC-2790 section of the MIB, it can cause some communications outages.
In these scenarios we normally recommend disabling of the process table and create process monitors within the empire MIB so that the agent can send traps when a process goes down, rather than depend on polling from the manger. This is a more efficient setup.
Also keep in mind that log monitoring configurations can cause performance issues as well. Smarter usage of regext for a single log/monitor is much more effective than having multiple monitors for a single log.