loadrunner Assume that I need to do client side...

  • Thread starter NaveenKumar Namachivayam
  • Start date

  • perf-test.com need your contributions to build up a strong repository of performance engineering resources.

N

NaveenKumar Namachivayam

Guest
Assume that I need to do client side performance testing where the application is built on ReactJS framework (e.g. Facebook, Netflix). We have TruClient protocol in HPE LoadRunner where I can use profiler option to see the statistics, as well as NV feature in LR 12.50 which is helpful in capturing client-side metrics. If I want to measure the performance of server side, may I use the same script or I need to convert that into Web HTTP/HTML script. The issue is: using the TruClient script, I cannot inject the more load. Also, if I use 3 LG, the response time is not within SLA, but if I use 10 LGs, the response time is better. Thoughts, please.
 
Sort by date Sort by votes
You already answer your own question. By adding more LG, response time become better means that your LG is the bottleneck
 
Upvote 0 Downvote
Use 1 user only in truclient. The other user should use normal web/http protocol. You dont need all user in truclient
 
Upvote 0 Downvote
Do not wait for performance testing (inherently a server scalability item) to measure client side performance. You need to have clear measurements and criteria for passing for unit, component assembly and functional testing. The developer tools in your browser can measure client side actions to the millisecond. Go into your server testing with your client performance as a known constant. Otherwise you get a an X+Y=Too_Long, solve for either X or X, where X is client impact and Y is server impact. Pair a handful of TruClient or GUI with the majority of transport level users. Be explicit in your transaction naming, so you have context such as Login and Login_GUI depending upon your user type. You can use this structure to see the weight of the client code under load, Login_GUI-Login=Client weight for Login.
 
Upvote 0 Downvote
Thanks James Pulley. Do you want me to test with only one vuser for server side testing to compare it with GUI. As I said, LG is a bottleneck, if I inject more than 4 vusers in TruClient protocol.
 
Upvote 0 Downvote
How about you place 49 http users on one load generator and 1 truclient on a second generator, by itself. Now you can measure both the server and the imposed client overhead
 
Upvote 0 Downvote
James Pulley Will the comparison valid? We are not comparing Apples to Apples, isn't it as the number of vusers are different.
 
Upvote 0 Downvote
Your server sees 50 sessions. It does not know or care if these sessions are created by http, truclient, gui virtual users or a collection of hyper caffeinated and pizza buzzed gnomes sitting at actual computers recording timing with stopwatches.
 
Upvote 0 Downvote
You must log in or register to reply here.
Top Bottom
Back