03-18-2014 12:01 PM
We just started using the TruClient IE protocol using 11.52 and IE10, and are seeing much longer transaction times in the TruClient protocol than for the same transaction run in the Web - HTTP/HTML protocol.
For instance, I scripted up a simple Login script using both Web - HTTP/HTML and TruClient IE protocols. The Login transaction occurs in about 2 seconds using the Web - HTTP/HTML protocol, but the same Login transaction takes over 6 seconds when run using the TruClient IE protocol.
Is there something we need to do to make the transactions run using TruClient run in the same time as in the Web-HTTP/HTML protocol? We've been monitoring the Load Generator PCs, and those aren't being overloaded and creating a bottleneck (we're running only 5 VUsers in the comparison tests, and have 10 Load Generator PCs available).
At this point, I'm unable to trust that the transaction times for any TruClient load test are accurate, so I hope to find out why asap.
03-19-2014 12:51 AM
Web/HTTP and TruClient are different protocols. The transactions in TruClient can measure other things than network round-trip as in Web/HTTP.
You need to specify the End Event of the transaction in TruClient so it will measure the desired time.
Additionally, the mechanisms are different. While web/HTTP measures the raw network level, TruClient is much more closer to the user and coupled with the browser, times may be different after all.
Hope it is clearer.
TruClient RnD Team Manager
03-19-2014 01:06 PM
Thanks for the reply, Gary, but I'm afraid it isn't any clearer.
I'm not sure what you mean by specifying the End Event of the transaction. I added both the Start Transaction and the End Transaction around the Login step. Is there something else I should be doing?
Based on which protocol I use, the number of logins per minute is different by a factor of five. For 50 VUsers, the Web - HTTP/HTML protocol shows 500 Passed/minute, but for TruClient IE I only get 100 Passed per minute. This makes it incredibly difficult to calculate the number of users I need to reach the expected load, which for this application is 15,000 logins per hour. With the Web - HTTP/HTML protocol, I'm already at two times the full expected load with 50 VUsers, but with the TruClient test, I'm at just over a third of the expected load with 50 VUsers.
How do I reconcile that the transaction times are so far off from each other? Have all of my previous Web - HTTP/HTML tests been that far off? That would be really unsettling to me, because I've signed off on applications based on data that may be off by a factor of five, if what you're saying about TruClient being much closer to an actual user is correct.
03-20-2014 01:40 AM
I think you are missing some knowledge on TruClient.
Are you familiar with the Transactions concept in TruClient? It is a bit different than in other protocols.
I just can't explain such wide domain in this meduim. If you are still lost contact me privately.
TruClient RnD Team Manager
03-31-2014 10:33 AM
First, everything Guy said.
Second, remember that the Web protocol is just sending data back to the Web server. There is no real client or browser. With TruClient, LoadRunner is creating a real albeit headless instantiation of a browser. There's quite a bit more overhead wiyth that, and depending on what end even you choose, you may add various levels of client-side work to the measurements.