Disclaimer: I am trying to get a feeling of real world numbers from peoples first hand experience. No implementation details / code is needed at this point.
We have web api that will use queues / message brokers to drop messages to and then in tu, windows services will read queues for delivering different kind of http notifications (simply: send many http requests to different urls/hosts). At peek times, those numbers might be in thousands per second.
Having on mind ServicePoint max coections limit plus ThreadPool threads, i am trying to find right measure for finetuning that sending part. So, potential questions are:
- What numbers of concurrent http requests (most probably through HttpClient) sounds reasonable to have per one windows service?
- What are some maximums that you use in projects without problems?
- If there are some specific things to pay attention to in order to accomplish the high number, please specify them.
Notes:
-
For the moment, i am not conceed about queue/message broker limits. that part we can assume is infinite for the sake of this discussion.
-
The requests are very lightweight with mostly a few kbs of data in them so response times are expected to be very short