Forum Discussion
AlexKaras
15 years agoCommunity Hero
Hi Johny,
The 500 status code not always means that the server is in non-operatable state (see e.g. http://support.microsoft.com/kb/311766). In this case you should either follow the recommendations from the referenced link or examine your web server logs to find out what caused server to fail to process the request.
As for all your other questions, I'd like to note, that despite e.g. functional testing, when you verify the deterministic result, in case of load testing you are rather watch how the server behaves under different load and then make a decision based on the performance figures *and business needs*. (We are not talking about crash of web server or tested web application -- if this happens then this is definitely a problem that must be addressed by Development.) In general, web application / web server does not crash when the load increases, but starts to respond more slowly or it may fail some requests, or send orphaned data to clients, or... And this is up to you and your business analysts / marketing to decide what level of performance slow-down is acceptable.
For example, the slow-down from 1 minute to 3 minutes when the load increases from one user to two users for the accounting report that is prepared once every six months by two accountants only, most probably is not a problem.
But it may be unacceptable if the slow-down increases from 3 seconds to 8 seconds for the online auction when the load grows from 50 users to 300 users.
Summarizing: load testing provides you with the figures of how responsive the site is under certain load. Then it is your part of work to figure out the reason of the failed requests (this may be because developers fail to handle concurrent requests / access to resources (like database access) or because web server needs more resources (CPU, memory, disk speed, etc.) ) and make a decision about the lowest acceptable response time for the typical and peak load.
The 500 status code not always means that the server is in non-operatable state (see e.g. http://support.microsoft.com/kb/311766). In this case you should either follow the recommendations from the referenced link or examine your web server logs to find out what caused server to fail to process the request.
As for all your other questions, I'd like to note, that despite e.g. functional testing, when you verify the deterministic result, in case of load testing you are rather watch how the server behaves under different load and then make a decision based on the performance figures *and business needs*. (We are not talking about crash of web server or tested web application -- if this happens then this is definitely a problem that must be addressed by Development.) In general, web application / web server does not crash when the load increases, but starts to respond more slowly or it may fail some requests, or send orphaned data to clients, or... And this is up to you and your business analysts / marketing to decide what level of performance slow-down is acceptable.
For example, the slow-down from 1 minute to 3 minutes when the load increases from one user to two users for the accounting report that is prepared once every six months by two accountants only, most probably is not a problem.
But it may be unacceptable if the slow-down increases from 3 seconds to 8 seconds for the online auction when the load grows from 50 users to 300 users.
Summarizing: load testing provides you with the figures of how responsive the site is under certain load. Then it is your part of work to figure out the reason of the failed requests (this may be because developers fail to handle concurrent requests / access to resources (like database access) or because web server needs more resources (CPU, memory, disk speed, etc.) ) and make a decision about the lowest acceptable response time for the typical and peak load.