TLDR
There are no good or bad policies for Autoscaling, and there are effective and ineffective for performing some functionality. CPU-based metrics may not effectively reflect the real load, as relation between CPU utilization and the number of requests/users may be non-linear. Concurrency based metrics may make you resource usage more effective. RequestPerTarget is a good metric to address steadily increasing and decreasing load. As for spikes, it may be a little bit slow to react, and as a result, may cause unnecessary downtimes and errors.via the TL;DR App
no story
