2 Replies Latest reply on May 1, 2020 10:48 AM by Scott Crawford

    Recommendation: Fewer Job Servers with Higher JVM Heap Sizes for Same Cost?

    Scott Crawford
      Share This:

      As we continue the development of our future cloud (AWS) BladeLogic/Truesight environment, we originally planned on 8 M5.xlarge job server instances.  However, going with 6 R5.xlarge instances will allow for 64GB more total memory (for the same cost), which seems to be the higher utilized resource in our current environment:

       

      R5.xlarge --> 6 * 4 CPU X 32GB RAM (192GB RAM Total)

      M5.xlarge --> 8 * 4 CPU X 16GB RAM (128GB RAM Total)

       

      Currently we have 8GB JVMs with the following parameters.  The majority of our jobs that get run are NSH script and deploy (file and BLpackage) jobs:

      MaxWorkItemThreads: 100

      MaxWorkerThreads:50

      MaxJobs: 50

      MaxJobThreads:5

       

      However, by going to a larger amount of total memory per instance, we were considering increasing the JVMs to 24GB each, and then increasing the number of threads available.  Does this seem like a good approach, and if so, what blasadmin job settings would best be best to adjust?