Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..
547 views

truclient: difference between Firefox and Chromium

Hi all, 

I have a question about TruClient Chromium: I haven't used this protocol before, but I see a lot of difference in response times (and performance in general) between two runs, one with Firefox and one other with Chromium: in the script recorded with Chromium there's a very decay and I can't explain that. 

Maybe Chromium (in incognito mode) affects performance in terms of CPU/memory more than Firefox? 

Thanks in advance for any type of suggestions,

Lorenzo

0 Likes
18 Replies
Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..

Re: truclient: difference between Firefox and Chromium

In add, here's a simple comparision between two run with the same scenario characteristics (50 Vusers, distribuited well around load generators -no more than 3 Vusers per machine-, duration 60 minutes): in the first we have Firefox's results (which are good for application in Production environment) and above the table about the same launch with TruClient-Chromium. 

lb1_0-1580116037803.png

Can anyone explain this difference?

Thanks in advance,

Lorenzo

0 Likes
Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..

Re: truclient: difference between Firefox and Chromium

That's the situation in a load generator machine: I'm explore the process ereditary with ProcessExplorer and I see two chrome.exe connection in a single node. 

lb1_0-1580117639322.png

Is it that a correct behavior?

Lorenzo

 

0 Likes
Highlighted
Acclaimed Contributor.. Acclaimed Contributor..
Acclaimed Contributor..

Re: truclient: difference between Firefox and Chromium

@lb1, Hi Lorenzo,

I see that you have a high failure rate with your Chromium test for one transaction. I think that you should fix that first before you can make any statement about differences.

Tip: when you are concerned about system resources. Run on each LG perfmon and monitor system resources including individual processes.

Success,

Erik

Signature:
Reward community members who take time to respond and help.
Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..

Re: truclient: difference between Firefox and Chromium

Hi Erik,

thanks for the response. As I wrote before, the two scripts are identical, the errors with TruClient-Chromium are caused by objects or strings not found. The application is going well, there's no alterations, and the distribution along load generators doesn't cause problem, in fact I haven't messages about CPU or memory leak but only for object not found,

lb1_0-1580196505568.png

Attached to this message, the comparision between perfmon (thanks for the suggestion!) with Firefox and Chromium. 

The problem, in a mine sensation, is on this approach with Chromium. I discover that there's three different chrome.exe processes per Vuser on each load generator (or localhost) and two prompt routines during a run, instead of one firefox.exe or TcWebIELauncher.exe per Vuser in other TruClient ways. 

Have you ever come across this problem? Because I have to justify the degradation of response times and failures with this Chromium...

Thanks again!

Lorenzo

0 Likes
Highlighted
Acclaimed Contributor.. Acclaimed Contributor..
Acclaimed Contributor..

Re: truclient: difference between Firefox and Chromium

Lorenzo ( @lb1 ),

In respect to the number of processes, Chrome is a little extreme. When I've chrome open with one tab, I've one parent and 10 child processes.

FF with one tab, has one parent and 4 child processes. In the old days FF did run (before version 40?) did not run different processes, but multiple threads. The multi process approach should bring more stability to the application.

When you open (procexp) one of those Chrome instances (properties) and go Performance tab, you might discover that most Working Set memory is Shared/Shareable and only a small set is Private.

I expect the difference in behavior between FF and Chrome is not the cause about the difference you observe.

I think that issues in respect to 'Object' not found are important in TruClient. Most of those wait actions have a timeout that needs to be elapsed before the conclusion 'Object not found' can be drawn. So that might be the cause for lower hit rate etc. Despite the effort of Micro Focus, it is not always possible to switch browsers without any effort. You should keep in mind that webservers might be the cause for different behavior. They might change logic based on browser type.

Success, Erik

 

 

Signature:
Reward community members who take time to respond and help.
Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..

Re: truclient: difference between Firefox and Chromium

Thanks for your interest!

Now, I know that different browsers bring different behavior, but this 2 seconds of difference need to be justified. In add, the two script in comparision are the same navigation, I haven't done the "translation" from VuGen but recorded one from beginning. 

The other strange thing is that: if I try to navigate with my browser Google Chrome during the load test, the application goes well (I tried often to take an error...but without success!), so I can't explain this behavior. 

What you think about that? Don't you think it can be a problem of this "heavy" TruClient-Chromium? As you saw in previous image, TruClient-Chromium has three processes and two prompt routines per Vuser, instead of TruClient-Firefox/IE which have one process per Vuser.

Thanks a lot,

Lorenzo

0 Likes
Highlighted
Acclaimed Contributor.. Acclaimed Contributor..
Acclaimed Contributor..

Re: truclient: difference between Firefox and Chromium

Lorenzo,

Did you try to navigate with the same browser as LR is using (Chromium) i.s.o. Google Chrome? You can find the execution path with procexp.

In this kind of situations is comes down to the tiny details.

Success, Erik

Signature:
Reward community members who take time to respond and help.
0 Likes
Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..

Re: truclient: difference between Firefox and Chromium

Hi Erik,

thank you again. I tried to use the Chrome version of LoadRunner, but, as I already said, the behavior is not the same as data results from running: Analysis and Controller told me that average time and TPS aren't good and there're some errors...instead, if I try more times a manual navigation with the browser (official Google Chrome or the version of LoadRunner) I noticed that there aren't errors and response time are good. 

I can't justify that...and it will be a problem. I didn't have the same issue with Firefox.

Lorenzo

0 Likes
Highlighted
Acclaimed Contributor.. Acclaimed Contributor..
Acclaimed Contributor..

Re: truclient: difference between Firefox and Chromium

Lorenzo, I'm running out of options. The last two things I can think of are:

- Run the TruClient script in VuGen with Chromium and see if it brakes and correct the not-found objects. (Validate if that still runs fine in FF)

- Record the scenario completely in Chromium again and use that script to compare against FF.

Signature:
Reward community members who take time to respond and help.
Highlighted
Outstanding Contributor.. Outstanding Contributor..
Outstanding Contributor..

Re: truclient: difference between Firefox and Chromium

Hi Erik,

thank you again for your time and sorry for the insistence. 

- Run the TruClient script in VuGen with Chromium and see if it brakes and correct the not-found objects. (Validate if that still runs fine in FF)

I've already recorded and run the script both in TruClient-Chromium and Firefox: in the first case, if I run it in silent mode (so from VuGen, but also in Develop Mode) the execution goes well, as for Firefox. I think Chromium is more "heavy" than Firefox so, in case of massive test, the Vuser load affects results, in fact it opens three chrome.exe process and two prompt routines per single Vuser...but I would like a MicroFocus technician approval.

- Record the scenario completely in Chromium again and use that script to compare against FF.

The comparision between Chromium and Firefox was already published in this thread (an image attached to this response), and shows bad results for Chromium (in terms of response time, TPS/TPH, passed/failed), maybe for the reason explained before. 

Thank you again, Erik. I hope for a justification about this difference. 

Lorenzo

0 Likes
Highlighted
Acclaimed Contributor.. Acclaimed Contributor..
Acclaimed Contributor..

Re: truclient: difference between Firefox and Chromium

Lorenzo ( @lb1 ),

in case of massive test, the Vuser load affects results, in fact it opens three chrome.exe process and two prompt routines per single Vuser

I would not focus on the amount of processes. Key for resource restrictions is limitation on CPU, Memory and Network for a LG. You represented two perfmon graphs one of FF and one of Chromium. The FF one has consistent higher CPU than the other (but the graphs only show 8 min, so not sure how to relate this to the overall throughput graphs).

I miss something in the throughput graph. The FF one has a High red line and the Chr. one a high and dropping yellow line. Do they both relate to the same rate or is the yellow line the error-rate and the red one the successful rate?

Can you run a longer test and scale up with 5 or 10 users per 30 min? That might show at which concurrency level Chr. test is getting (more) errors. Also record a perfmon profile (LR can do that as well for you) the system resources of the LG (CPU, Available memory, NW).

Success, Erik

Signature:
Reward community members who take time to respond and help.
0 Likes
The opinions expressed above are the personal opinions of the authors, not of Micro Focus. By using this site, you accept the Terms of Use and Rules of Participation. Certain versions of content ("Material") accessible here may contain branding from Hewlett-Packard Company (now HP Inc.) and Hewlett Packard Enterprise Company. As of September 1, 2017, the Material is now offered by Micro Focus, a separately owned and operated company. Any reference to the HP and Hewlett Packard Enterprise/HPE marks is historical in nature, and the HP and Hewlett Packard Enterprise/HPE marks are the property of their respective owners.