My web server has a lot of dependencies for sending back data, when it gets a request. I am testing one of these dependency applications within the web server. the application is decoupled from the main web server, and only queries are going to it in the form of API’s exposed.
My question is, if I wish to check these API’s in a multi-threaded environment (C++ functions with a 2 quadcore processor machine), what is the best way to go about doing it?
Do I call each API in a separate thread or process? If so, how do I implement such code? From what I can figure out, I would be duplicating the functioning of the web server, but I can find no other better way to figure out the performance improvements given by that component alone.
It depends on whether your app deails with data that’s shared if it is run in parallel processes because that’ll most likely determine where the speed bottleneck awaits.
E.g, if the app accesses a database or disk files, you’ll probably have to simulate multiple threads/processes querying the app in order to see how they get along with each other, i.e. whether they have to wait for each other while accessing the shared resource.
But if the app only does some internal calculation, all by its own, then it may scale well, as long as all its data fits into memory (i.e. not virtual memory access, e.g. disk access, necessary). Then you can test the performance of just one instance and focus on optimizing its speed.
It also might help to state the OS you’re planning to use. Mac OS X offers tools for performance testing and optimization that Windows and Linux may not, and vice versa.