What actually increases the performance of an Appian environment. Is it increasi

What actually increases the performance of an Appian environment. Is it increasing the number of kdb servers with process execution engines tuned up or increasing the number of app servers. I have a requirement to increase the performance of an environment that is performing deadly slow. Currently it has 1 kdb server with all engines and 2 app servers with 1 searchserver located on one of these app server. Apache is also running on the app server. Increasing the number of app servers is much unlikely due to redhat licenses. Actually we paid 33k for adding an extra app server recently. So will it matter if we increase the process execution engines or any other ideas are highly welcome. We saw an increase in the number of java classes with a monitoring tool when we upgraded our system from 7.9 to 7.10. In the morning to 2pm hours its at about 80k classes and sometimes in the evening it jumps above 100k and then crashes the server. Any i...

OriginalPostID-200039

OriginalPostID-200039

  Discussion posts and replies are publicly visible

  • ...deas to fix this issue are highly welcome
  • I'd recommend debugging as follows:
    a) Reproduce on a non-production environment
    b) Remove customizations and custom java code
    c) Located offending custom code or plug-in through testing
    If the above isn't followed, you will likely continue to have problems. Need to get to the source of the problem and missing controls that introduced the issue.
  • Follow the previous advice to get to a stable, testable state.

    Then use Appian's performance logs to identify whether there's a bottleneck in your app server or your engines (or both): forum.appian.com/.../Logging.html

    With that knowledge you can determine where you may want to scale your architecture: forum.appian.com/.../Scaling_Appian.html
  • If the platform itself is running slow, perhaps you could identify issues with the applications that are causing the slowdown. Things like Data Management of process models (Deleting after 1-3 days), thinning your process instances by making use of small, more modular process models with asynchronous sub-processes and termination end nodes, deleting any un-needed large documents in the knowledge centers, avoiding constant import/exports of large applications, etc might help the environment. Fixing the root cause is always better than putting on a Band-Aid if possible.
  • Unfortunately this application has large documents...some of them several MB stored inside within Appian. Process Models are huge with lot of nodes and most of the general common functionalities have been broken down to sub-processes which are also huge....They also use custom built-in plugins for integrating to external systems using web services.....this application actually talks with more than 10 external systems that have additional security layer built in and this application itself is on single-sign on as well.....but the good thing is that not that many people are using it at the same time(may be 200).....so I do not know where the bottleneck is that is causing the root cause of slowness as there are so many parameters........i don't think it is possible for us to create a non-production environment and simulate this on something like jmeter because of so many layers of authentication and bridges that has to be passed through.....but will it be helpful in any manner if we scale up the process-exec or some other engines to separate servers to increase the performance in any manner....if jboss has to be scaled i have to give solid evidence to my manager and director that we need to scale up more which is something that they don't want as they recently spend like 33k on red hat licenses....