Support for token bucket - enable burst quota
The current quota/call rate limit is +1 per call. In practice this means we create SKUs based on the maximum expected spike rather than average usage. By supporting a token bucket model (https://en.wikipedia.org/wiki/Token_bucket) we could define a SKU more aligned with our actual usage.
For example: on average we have 50 calls per second, but need to be able to spike to 250 calls per second.
Today we'd create a 250 calls per second throttle policy for this key/product which is not optimal.
1
vote
Ryan Galgon
shared this idea