Spike Arrest central counter

In Quota Policy where we have a central counter which keeps track of API counts and tells message processors about the actual quota count, but In case of Spike Arrest do we have any logic to keep the count and divide each request into particular time like if we have 4 counts per minute then how does Spike Arrest policy divides 1 call per 15 seconds and how does the actual counts are being updated and where?

0 1 84
1 REPLY 1

If I'm not mistaken, 4 calls per min, will allow only 1 call in "border" of 15 seconds, second call inside 15 seconds window will be rejected.

Counters stored in Cassandra in counters keyspace, and exists in L1 cache on MP level, I think