What are possible causes for sudden decrease in throughput when load test enters a steady state ?
I am running a test with a load of 250 users and I can see sudden decrease in throughput once all users are ramped up.
What do the graphs say? have you merged the throughput graph with CPU / Mem usage graphs?
usually a sudden drop in a KPI means a bottle neck somewhere else.
review the AUT log files, look at server monitoring graphs, try with a lower number of Vusers and find the bottle neck.
This is the fun part of performance engineering :-)
Remember some key rules regarding Throughput:
1. if the throughput scales upward as time progresses and the number of Vusers increase, this indicates that the server processes the request without any issue as well as the bandwidth is sufficient.
2. If the graph were to remain relatively flat as the number of Vusers increased, it would be reasonable to conclude that the bandwidth is constraining the volume of data delivered, no issue with server performance
3. (Most likely your case) If a sudden drop was seen in throughput during ramp-up or steady state then it indicates that network bandwidth is sufficient, but something went wrong with the server performance.. In that case, you need to look the server side metrics like CPU and Memory Utilization of the server.
Solution: Merge throughput graph with average response time, CPU Utilization and Memory Utilization graphs and find out the root cause.