Closed samlaf closed 3 weeks ago
Continuing this thread - how do we determine a customer's rate limit without having them specifying it manually? CC - @samlaf, @bxue-l2, @teddyknox
Continuing this thread - how do we determine a customer's rate limit without having them specifying it manually? CC - @samlaf, @bxue-l2, @teddyknox
Honestly rethinking about this.. is it really needed? Feels like a simple exponential backoff should be enough? Prob not worth overcomplicating things unless absolutely needed?
Closing as not planned. This is too complicated and not needed, at least for now. Can reopen if some customer is interested.
Creating this issue to discuss where is the best place to handle rate limits and backpressure in general.
With new op-batcher that supports concurrent dispersal requests (see https://github.com/Layr-Labs/optimism-exp/pull/3), we need to start considering how to make sure not to go over the disperser rate-limits if multiple blob requests are being sent in parallel.
We could add a rate limiter in the op-batcher code, but wondering if it wouldn't be best to add the rate limiter as part of the eigenda-proxy middleware. Probably could allow a buffer of requests to wait before dropping requests to put backpressure.
Pros:
Cons: