Hi All,
We're executing load tests on our testing environment. We notice that during the high peak, simple operations like write or query records take longer to execute (2-4 sec).
The CPU memory utilization of the instance is very high too.
The records are light weighted and not heavy to take this long.
Any ideas as to what might be the cause of this slowness and how to debug this?
Regards,
Sakshee
Discussion posts and replies are publicly visible
That is expected. See, when Appian initiates a node in a process, the internal implementation is not started immediately, but added to a queue. So the actual duration is typically an order a magnitudes lower that the number you see in the process.
What kind of load testing are you doing? Are you doing a lot of writes? A lot of queries? Or other operations?
We're doing a bunch of write record, query records and a couple of API calls as well in the end to end flow. The load tests were carried out with 3 concurrent requests and upto 150 requests in an hour. With this load we see slow record write and get query performances (2~4 sec each).
I think it's hard to diagnose this over chat without a lot more information. In general I would suspect if you have several concurrent queries and writes that it will be slightly slower than without the concurrency. If you're seeing results that concern you, I'd suggest opening a support case.
saksheer5952 Were you able to figure out these issues, we are experiencing similar slowness while running performance tests.
The data fabric doesn't have a way to compartmentalize / ensure quality of service for users over non-users, or even one app over another. It would be nice if we could have priorities for queries, but those would mask a larger scalability problem in how you've setup your system.First, run through the 'Scaling Appian' page and triple-check the results on your performance testing environment. https://docs.appian.com/suite/help/25.2/Scaling_Appian.htmlIf your architecture/hardware seem to be correctly sized and configured, there are some additional questions to ask/answer about how you're using the data fabric. Could you elaborate on the frequency and types of calls made to the data fabric? For example, are your performance tests UI/user-based, or are they API-based?