Camunda Self Hosted performance tuning

Camunda 8 Topics Discussion & Questions
1

Hi there.

I’m baking POC project with usage of Camunda 8 (self hosted managed by helm charts).

There is my configuration of zeebe.

clusterSize: “3”
partitionCount: “9”
replicationFactor: “3”
cpuThreadCount: 15
ioThreadCount: 15

resources:
    requests:
      cpu: 1000m
      memory: 6000Mi
    limits:
      cpu: 1000m
      memory: 6000Mi

I’m running 1000 processes loaded once in time (simulation of 10 concurrent users that start 100 processes each) to get the understanding of throughput, latency, general performance.
Process is a simple queue of 5 service tasks that makes REST GET to a resource with response time 1000ms invoked by webhook.

Screenshot 2024-05-14 at 13.49.31
Screenshot 2024-05-14 at 13.49.311878×284 18.1 KB

In result I have all processes completed in approximately 8 minutes 30 seconds.

Nor clusterSize, partitionCount, cpuThreadCount, ioThreadCount, cpu, memory changes have improved the performance of process throughput.

Could anyone please helm me to tune Camunda cluster properly ?
Which parameters should improve the performance ?

Thanks in advance.

2

Hi, problem lies in cpu limits. By 1000m you set, that only one CPU can be used. Also not sure about rate of cpuThreadCount to ioThreadCount. Its something I wanted to test, because I saw distinguished configuration. But I think, that thread count should be more to cpuThread instead of ioThread. If you will test it, please let us know.