API QOS Layer : System wide limits

System level limits should be imposed and configurable. A rate exceeded error should be returned to all requests while these limits are exceeded. The actual system load should be considered as well as gross requests per second.

Defaults:
System load average over 50% ( since this measurement is averaged over the last minute, the system could be at 100% load for 30 seconds before this limit kicks in at 50% )
500 requests per second

These requests per second limit should be configurable