I am testing a website using LR to find the number of users the website can scale within the threshold. I am testing with 1000 concurrent users while ramping 10 users for every 3 minutes. I have defined the SLA's for different transactions.
Can the LoadRunner Report suggest how many users the website can hold based on the SLA's defined.
I know that there is a graph for "Transaction response time under load". I don't want to filter the graph since I am using the 90 percentile response time.
Can the LR do a self analysis and suggest the same.
Create a step wise workload with 200, 400, 600...1000. Even during execution you can see clearly for what load the site is experiencing issues. Just a merge of running vusers vs response time graph or TPS will clearly indicate the bottleneck period.
Thanks for your reply.
I did stepwise workload for 10 users upto 1000 users. There are many transactions. The SLA for different transactions are not the same. I guess, i have to filter by time for each stepwise workload and verify the SLA's have passed. However, i got your point.
I have a followup question, kindly advise
I need to test the website for 1500 Virtual users. I have Virtual user licenses for 1000 only. If i reduce the think time/pacing to match the number of iterations that a 1500 virtual users.
a) Will it be appropriate to compare the results with 1000 vusers with a load of 1500 vusers and 1000 virtual users with normal load.
b) Also, How would i specify the number of Vusers in my report for the 2 performance test executed? since my 2nd test would be also 1000 vusers but a load of 1500 vusers
I think moderators would prefer you post new questions in a new thread.
This has been discussed many times on the forums. The only way to test 1500 vusers is to run a test for 1500 vusers. Just increasing the transaction rate of 1000 vusers is not the same thing as running 1500 vusers (it is 1000 vusers working harder). If you have a problem with database connections and it doesn't show up until you have 1300 concurrent vusers, then your "1000 vusers running faster" test is not going to expose this problem. That's one example of many. My advice is to make a strong case with management to purchase the additional 500 vusers (there are other options besides permanent licensing that is less expensive).
Disclaimer: I do not work for HP, nor do I make any profit from your purchase.