TestComplete and TestExecute performance problem

Everything works fine after reinstallation of Windows OS and TestComplete 9.31 on physical machines. According to 'Super PI Ver1.1e (calculation of pi up to 33.55 million digits)' benchmark there is almost no difference in performance of described machines.



Currently the issue exists on another virtual machine with Windows Server 2008 R2 Standard. I installed on it a trial version of TestComplete (ver. 10.0.531.7). The result is very simillar and scripts execute 7 times slower than on physical machines. I have copied whole project directory to this virtual machine from physical machine.

Today the same issue has appeared on two physical machines mentioned above. This is very strange that at the same time two instances of TestComplete (ver. 9.31) on two independent machines have slowed down. We don't know what can be the root cause of this issue. The same problem was observed some time ago on one physical machine (first post above).



I have no idea what is wrong but as far as I know there wasn't any change in network connections as well as configuration and settings on both physical machines remain unchanged.



Does anyone have a similar problem?

Michal,



I've been wrestling with the same problem for the last couple of months VMWare virtual machines running distributed tests under Windows 7.  One machine always runs about 3 times slower than the other two. It's interesting since, they're all clones of the slow machine. Rearranging jobs among the machines does not change the performance. Lately, a second VM has slowed down also. Restarting the the VM's seems to have some positve effect, as does restarting TestExecute and the application under test. 



Suspecting that the cause could be a bad VM build,  the machines were rebuilt and reconfigured with more physical processors and memory. This helped, but the difference still remains.



Let me know what you find out.

Mark,



I have updated one of the physical machine to TestComplete 10 but nothing has changed. Delays in test execution are the same as on TestComplete 9.31. I expected that something can change after update but it didn't. Moreover I observe situation is worse with every hour - delays are bigger and bigger. For example the same actions which were executed below 20 seconds in Firefox browser before problem was noticed (at about 10 AM local time), now are taken over 7 minutes.

We have noticed in 9.31 TestComplete and the version we used before (8.x I believe) that during the day on any given machine, TC will slow down and also stop recognizing objects that have not changed in the application under test.  We usually have to restart the application under test and TC to fix this and sometimes we have to reboot the machine.  



Happy to know we are not alone in this but I would be happier if there was a solution!

Sorry it took a while to respond...



The slowdown for me at least has always been a problem, for as long as I have used Test Complete. We are currently using TC 8.7, and it is stable enough that I don't dare venture forward. I have used TC9, and 9.1 in the past. I did not try TC10, and I couldn't anyways since there is no way to convert my TC8.7 scripts without having to do loads of changes (and in a ~4500 tests project, that is just not realistic). There are severe memory leaks and performance issues in TestComplete, at least for all of the versions I have used. Some are better than others, but the bottomline is if it doesn't crash you should probably consider yourself happy.



Others and I have pointed out many performance issues that still remain unresolved to this day. Smartbear is more interested in adding features and bloat than solving these issues in the long run it seems. It's probably just easier.



In any case, on VMs there are TONS of things that can be misconfigured without you even knowing about them. Having the slowdown issue on a physical host however is weird. Two identical machines started at the same time should pretty much take the same amount of time to complete the same tests. If one is significantly slower than the other, but proper benchmarks see no difference in performance, there is a good chance some software or service is interfering, and often that is an anti-virus or something similar. As a rule of thumb, your test machine or VM should be bare bones. That means disable every non-essential service (like Windows Search, or System Restore, Compatibility Assistant, Windows Firewall, etc...), clear up any non-essential third party sofware, including anti-virus or other resident program, and make sure there are no connectivity issues. If after all that there is still a large difference in performance between the two machines, it might be a BIOS setting (especially power management like SpeedStep... also pay attention to Intel VT options in there...).



You can expect TestComplete to be slow and sluggish after a while. That is why in my case I kill the process completely after each test and start it fresh.