Approaches for Rate limiting for Feature flag service [Design doc]
Shubham Yadav, Vikhyat
We wanted to limit our AWS costs after a certain amount of requests, say the allowed request count is 500, we should not incur the costs.
- Things used in this solution are Dynamodb, one lambda for changing the config [To mark the reserved concurrency to zero]
- Steps in this solution
- When a request comes to FF service, we first check if the value stored in DDB table is greater than 0
- If yes we continue processing the request and return the requested API response
- If the value in the database has become 0 then we make call to the new lambda to make the
reserved concurrencyto 0 - This will make the API gateway not able to connect to the lambda
"Internal server error"
- Costs incurred in this solution
- Taking an example for 10000 requests made and the limit for the lambdas is 500 request
- So for 500 requests the lambda costs is $0.
- For rest 9500 requests the lambda costs is $0.04 rounding off to $0.5
- Now considering the dynamodb costs
- Dynamodb costs for 10000 requests is around $0.2, rounding this off to $0.5
- Data transfer costs is around $0.09 per GB rounding off to $1
- Considering the max data size = 10kb, so which means, for 100000 requests we will incur 0.09$ approx rounding it off to $1
- API gateway costs for first 1 million request = $3.5
- Taking an example for 10000 requests made and the limit for the lambdas is 500 request