What is the best tool (or tools) for monitoring the performance of the concurrent manager?

What is the best tool (or tools) for monitoring the performance of the concurrent manager? In particular we need to be able to identify long-running requets (at least based on a single figure; ideally based on a baseline for each program), a buildup of pending (standby and normal) requests (indicating that the concurrent managers are all busy and/or incompatibilities are causing issues), terminated and errored requests, etc.? Do people usually set up Alerts, use cronjobs that run periodic SQL, use Enterprise Manager, or use other tools?

We run around 20,000 requests per day, with around 200 distinct program names across 3 concurrent tier nodes. We have a fairly high degree of "incompatibilities" that restrict concurrency, especially around period-end, so actively monitoring the concurrent request throughput is a big issue.