Improving Rate limit for APIConnect Gateway cluster

Currently the rate limit across the DP GWs in a peering group(3 DPs in a peer group in the same Data Center/Availability zone) for APIConnect GW is working as expected for longer intervals. For example, if the interval is 10 seconds and the ratelimit is 100 then it works with intended accuracy. In case if the interval is very less like 3 seconds and the threshold is 100 transactions, then the rate limiting at APIConnect is not so efficient. Increasing the interval is not going to solve our situation where we want to restrict 100 transactions per 3 seconds.

We had a detailed discussion with IBM representatives about this and understood that the sampling across the peers(Primary and Secondary) should happen twice within the interval to reconcile for the rate limiting to work efficiently.

Now, our suggestion is that the technique or the way the rate limiting is being done currently has to be improved in such a way that it is still able to handle higher TPS for very short intervals. These are very essential for the APIs that are created to cater regulatory and compliance requirements set by the authorities.

CASE - TS009701873, we created this case to debug a different issue but later we also had a discussion about the above mentioned concept and issues as well over a webex call.