Question

I am implementing a REST API which requires throttling. I know that, ideally, you would place this logic in nginx. However, I have some unique constraints.

Namely, I have one class of users who should NOT be rate limited. It would not be useful to implement a rate limit on a per IP basis (the nginx way).

Users of the API are differentiated on a APIKey basis. Using a caching system, I could count requests per APIKEY and handle rate limiting accordingly. That involves more setup and is not as scalable, I would imagine.

Any suggestions?

Answer 1

You could setup multiple virtual hosts that are individually throttled at different limits. You could do your count and then redirect selected users to these virtual hosts to be throttled.