Closed joslat closed 8 months ago
@joslat Thanks for your feedback! We will investigate and update as appropriate.
Joslat, I'm following-up on this internally, and your feedback has been shared with the content owner for further review.
I am going to close this item as it sounds like more of a product performance issue than a documentation issue. If there is a change in limits doc that is needed, please let me know.
Dear App Service product team,
I have identified an issue with the Standard S1, S2 and Premium P1V2 instances, where on the connectivity it says unlimited web sockets and Ip connections are more than enough when I deploy a simple API endpoint to them (PI calculation, CPU intensive) and have them do like 10 requests per second for about 2 minutes, the app service behaves properly (5-10% CPU, 100-200Mb stable usage) and response time of up to 1, max 2 ms. But there are unresponsiveness peaks of 50 ms up to 700 ms, regularly, like a heart beat.
Like a connection bottleneck which should not be there according to the product specifications, so either they are wrong or there is a bug with them.
I have tested this with P2V2 and P3V2 and this does not seem to happen, not at 5-10% CPU nor at 60% CPU max, so it is not a throughput or performance issue.
I realized on this while preparing a conference talk for how to optimize, deploy and test a mission-critical API on Azure App Service Which I am improving and doing some variations of it, https://www.alldaydevops.com/addo-speakers/jos%C3%A9-luis-latorre-millas and https://sessionize.com/s/joslat/breaking-your-rest-apis-to-make-them-stronger/60141
I have the API and test code here too: https://github.com/joslat/AzureAPiPerformanceTesting And I'll be glad to clarify any point and help to make Microsoft's products better :)
Best, Jose Luis Latorre Microsoft MVP & K6 Champion
Document Details
⚠ Do not edit this section. It is required for learn.microsoft.com ➟ GitHub issue linking.