Slow page load time when increasing concurrent users

Question

I've been load testing our site recently and noticed that the browser page load time in Azure's Application Insights suffers a great deal when ramping up the number of concurrent users. The page scores highly in GTMetrix tests (96% performance, 76% structure) and loads within 2 seconds in isolation.

During the test, the average page load time increases to around 5/6 seconds but the 95th percentile can be from 10s up to 25s. I've checked the obvious things like CPU, Memory, Database etc in Azure and all of these are well below being maxed out.

The test which was run was 187 concurrent users (22 second wait time) over 4.5 minutes.

When looking into the slowest requests, most of them have messages like '47.40% of this request was spent in waiting'... This doesn't point at any particular piece of code, it's different per each request so potentially suggests some kind of blocking?

Is this potentially an issue with the number of requests being made? Or simply the efficiency of the code? Or something else...

I'm curious as to what can affect the load time of a page client side and what can be done to improve this with an increased load?

Answer 1

It might be the case of not proper configuration of your application server/database/other middleware, for example if the number of concurrent users is higher than the maximum number of connections in the pool - the remaining ones will "wait" until the next connection becomes available for serving them

So double check your application configuration and logs, there should be some clues there.

Another possible reason is that your code is not efficient enough like too complex algorithm is being used so it makes sense to repeat your test with profiler tool telemetry enabled and see where the application spends the majority of time

In general Azure App Insights is a very powerful APM tool which capable of providing the full information, it's just a matter of proper configuration/placement of the Agents