Forum Discussion

bpistole's avatar
bpistole
Contributor
9 years ago

LC 4 - huge variance in page load times between LC4, LC3 and IE

I am seeing a huge difference in the time it takes to load a page between LoadComplete and IE11.  I am using the same host, same client, same browser same log in credentials but still I see this huge difference. I've attached the Slow Pages (Average) requests as well as the request data from IE11 debug screen.  IE loads this page in 3-4 seconds while it takes about 17 seconds in loadcomeplete.  I've run this numerous times to see if it was an anomoly but it is not.  It is very reproducable, I get very consistant times.  I've even put LS in a loop and then manually executed the same steps and I still got the same results.

 

Compare to the two graphs to see what I am referring to.  There is a long gap in the LC load time while the manually executed one is nice and tight.

 

This is a single user logging on to a site and the landing page loading.  I'm stumped as to why I am getting this.

 

Thanks

  • For comparisons sake I've executed the exact same scenario on the same clinet, browsers, host, etc.  LC3s times are much more in line with what you would expect.

     

    Any help would be appreciated.  I'd really like to move to LC4 if I can get this resolved.

     

    Thanks

  • Hi Billy,

     

    So this is an interesting discussion that we have been having internally. Similar to my reply to your question about inserting new pages in LC4, this is also related to the concept of simulating parallel requests like a browser. There was actually a bug in LC3 that would start a new connection whenever you created a new page, which is not realistic. It did not have a huge impact on performance metrics, but we like realistic tests. :)

     

    In LC4, we implemented automatic "connection think time" to account for this discrepancy. If you go to the new Timeline tab for any page in the scenario editor, you'll see all of the connections and the wait times. You can manually adjust these by dragging the connections.

     

    However, we have already improved upon this mechanism to preserve realistic connection handling but also account for some logical inconsistencies in the connection think time approach. I've tested the new implementation and it's much closer to what you'd expect, as in LC3. We anticipate this update to LC4 will be delivered along with those other bug fixes in the next week or two.

     

    Best,

    • bpistole's avatar
      bpistole
      Contributor

      Thanks, I will looking connection think times you mentioned.  I thought the generic think time value account for this.  I appreciate the quick response.

      • RyanHeidorn's avatar
        RyanHeidorn
        Staff

        Sounds good. That's part of the discussion we've been having here at SmartBear... we now have (at least) three ways of controlling think time in your scenarios. At the page level (as in LC3), at the request level, and at the connection level (and not counting new operations like Delay). This is great in terms of flexibility, but makes it difficult when you want to control all timings at once. We are hoping to deliver an option to remove all think times in a scenario with a single click, which may or may not be coming in this next update. The other side of the argument is that by removing all think times, you could actually be generating more load than real users would, but my opinion is that it's easier to start from zero and add in think times that are logical to you (or set a random range for think time at the test level).