Question

I do not understand why application goes down when there are too many requests? How can I test my application for its maximum capacity to respond to a requests?

Answer 1

I do not understand why application goes down when there are too many requests?

If you are asking about a specific application, the answer is the we don’t know either, and we don’t have any evidence from which we could deduce what the problem might be. Places to look for evidence:

• Check server the logs.
• If you have a reverse proxy, check the proxy’s logs.
• Check the backend database logs.
• Check the system-level performance stats; e.g. look for evidence of VM thrashing, disc I/O overload and so on.
• Check for pathological GC behavior.
• Check for blocked / jammed threads, indicating concurrency bottlenecks, deadlocks and so on.

If you are asking in the general case, there can be any number of causes:

• Running out of an external resource (e.g. database connections)
• Creating too many threads
• Trying to queue every request to avoid dropping requests, and other broken strategies.
• Concurrency bottlenecks
• Pathological algorithm or database query design that results in performance that decays under load.
• Broken error recovery; e.g. underload you start getting unexpected exceptions, and this triggers resource leaks.
• And so on …

and how to test my application for its maximum capacity to respond to requests …

Create a test harness to repeatedly send requests to your server and see what rate the server can sustain. There are existing freameworks for doing this.