memory and CPU alerts when they get above/around 90%; dispatching caching rate, a general QPS overview and per-API-type latency metrics