This is mainly a performance issue. It is to help get the current average of the past N (30?) values. The context here would be to monitor the average execution time for the past N data requests in a server. While the obvious solution, is to log the all past N requests, in which all N request is read, and tabulated. Creating a performance hit relative to the data request in question.
As this is mainly use as a means of measure and not a perfect running estimate, the question? To solve this problem in the most effective manner.
While the solution may be language neutral, I would be implementing in php 🙂
Instead of saving the last N values, organize your values in “buckets” with a predefined bucket-size, and for each bucket, only save the sum of all values in the bucket.
Whenever a bucket is full, you can delete the oldest bucket. (The best way to do this would be a ring buffer)
This allows you to cut your memory usage by a factor of BUCKET_SIZE, but obviously your average is not over the last N values anymore, but over the last N to N + BUCKET_SIZE.