Turbonomic has been supporting over 20K entities in environment but with few number of targets. This article talks about when number of target is at large scale eg 600 and total number of entities is like 3000.
Below configuration is based on simulation 600 AWS targets in turbonomic lab. We have used modified utilProbe simulating exactly like AWS probe in term of latency and threads.
Recommended setting for turbonomic to support 600 pls targets
- Increase DiscoveryManagerImpl thread pool size to 50: change numThreads="10" to 50
- Tune post-processing to run once per 1 hour: add maxTimeWithoutPostprocSeconds="2400" and minTimeBetweenPostprocSeconds="1800" right after numThreads
- Configure market to run every hour: market.60min.config.topology will be provided
- Increase CPU and Memory: m5.2xlarge is recommended
- Increased disk space, 600 discoveries generate 1200 discovery dumps in /srv/tomcat/data/discoveries. Also 600 targets increase log size dramatically. + 30 additional Gb
AWS Probe changes
- in sdk.probe.properties adjust thread pool sizes