Best Way To Test Multi Users
Discussion
Hi Guys
Ive been asked to test the performance of a new system that being developed by my company and they want to simulate multi users, using the system at the same time. I and they have never done this before and they want suggestions on a program package that could do this role.
I have thought about loadrunner?
Any suggestions.
Cheers
Matt
Ive been asked to test the performance of a new system that being developed by my company and they want to simulate multi users, using the system at the same time. I and they have never done this before and they want suggestions on a program package that could do this role.
I have thought about loadrunner?
Any suggestions.
Cheers
Matt
Loadrunner is expensive. I think they do a free version but it's limited to 10 concurrent users (www.mercuryinteractive.com). If that's the expected load then it may well do the job
If its just a simple test required then I'd use webinject ... You script what you want it to do and it performs the tasks . You can use it to 'roughly' guage the performance ...
my 10p
If its just a simple test required then I'd use webinject ... You script what you want it to do and it performs the tasks . You can use it to 'roughly' guage the performance ...
my 10p
advice from god of lod testing (not me obviously)
----------------------------------------
Web:-
If it's web and your management want a "traffic light approach" then
something like webinject would be fine.
If it's a medium size system (several hundered concurrent users) I'd go for
Radview Webload, fairly cheap, easy to use etc.
If it's a large or complex system (over 1000) I'd use Loadrunner, it's the
market leader for a reason.
----------
Non Web:-
Use Loadrunner, get a Load Tester in to do it.
----------
A few other questions to ask, especially on a larger system.
Do you have a realistic amount of data in your db? (Nothing worse than
putting the prod load in and the system grinding to a halt - I've seen it
happen)
What are the performance requirements?
How many users are there / what are they doing?
Is your testing infrastructure the same power / size as your prod
infrastructure?
How are you going to fix any defects found?
----------------------------------------
Web:-
If it's web and your management want a "traffic light approach" then
something like webinject would be fine.
If it's a medium size system (several hundered concurrent users) I'd go for
Radview Webload, fairly cheap, easy to use etc.
If it's a large or complex system (over 1000) I'd use Loadrunner, it's the
market leader for a reason.
----------
Non Web:-
Use Loadrunner, get a Load Tester in to do it.
----------
A few other questions to ask, especially on a larger system.
Do you have a realistic amount of data in your db? (Nothing worse than
putting the prod load in and the system grinding to a halt - I've seen it
happen)
What are the performance requirements?
How many users are there / what are they doing?
Is your testing infrastructure the same power / size as your prod
infrastructure?
How are you going to fix any defects found?
Its a VB application - non web based.
The most users it should see at any time is about 30. Basically it is a system that tv stations can use to plan, transmit and sell airtime. So everything from commercials, promotions and programmes and they're contracts to billing and analysis.
Any defects found will be documented and passed to the dev guys who wrote the code for that area. Then retested by either myself or another tester.
They want something that can be used by existing staff, i have some exsposure to winrunner and loadrunner from a previous job but it was quite limited exsposure. Hence my question.
The most users it should see at any time is about 30. Basically it is a system that tv stations can use to plan, transmit and sell airtime. So everything from commercials, promotions and programmes and they're contracts to billing and analysis.
Any defects found will be documented and passed to the dev guys who wrote the code for that area. Then retested by either myself or another tester.
They want something that can be used by existing staff, i have some exsposure to winrunner and loadrunner from a previous job but it was quite limited exsposure. Hence my question.
Hi,
Try writing some test stubs within VB, you can then simulate various actions of a user such as opening the new customer screen, typing in a load of random data, saving it, doing lots of searches, etc.
You should in theory be able to test just about everything, just how you want it tested. Use your existing code to determine when certain actions have completed and timers to simulate a user pausing for a short time.
Subject to the load on the PC you could compile this test app and then run it 10 times and see what happens. If the PC can't keep up, run it on some more PCs.
You could be clever and add in some logging so it stores the time taken etc, but just by watching one of the test apps you can tell if its running very slow, such as the server can't keep up with searches etc. You can also test the input/output results by using pre-made sample sets, but the random way is pretty good for just load testing.
I did this with some management software by simulating 100,000 orders which were created over a couple of nights by running the 4 instances of the app at a time.
To avoid adding lots of test code to the real app, and risking breaking something, make a copy of the whole project then make the changes. No risk of releasing something with a bit of a test stub left in.
Sam
Try writing some test stubs within VB, you can then simulate various actions of a user such as opening the new customer screen, typing in a load of random data, saving it, doing lots of searches, etc.
You should in theory be able to test just about everything, just how you want it tested. Use your existing code to determine when certain actions have completed and timers to simulate a user pausing for a short time.
Subject to the load on the PC you could compile this test app and then run it 10 times and see what happens. If the PC can't keep up, run it on some more PCs.
You could be clever and add in some logging so it stores the time taken etc, but just by watching one of the test apps you can tell if its running very slow, such as the server can't keep up with searches etc. You can also test the input/output results by using pre-made sample sets, but the random way is pretty good for just load testing.
I did this with some management software by simulating 100,000 orders which were created over a couple of nights by running the 4 instances of the app at a time.
To avoid adding lots of test code to the real app, and risking breaking something, make a copy of the whole project then make the changes. No risk of releasing something with a bit of a test stub left in.
Sam
Gassing Station | Business | Top of Page | What's New | My Stuff