The following benchmark has been compiled on an Amazon
instance, which has 2 CPU cores and 8 GB of memory available.
CPU benchmarking was done using
CPU usage guidelines
Benchmark - CPU usage for particular average message size and EPS
Measured CPU usage for particular Events Per Second (EPS) average message size.
Benchmark - EPS for average message size and CPU usage
Events Per Second (EPS) achieved for a particular average message size and CPU usage.
|Average CPU usage||-||-||-||-||-|
Using the information from the above table, if you had an average CPU usage of 5%:
- 10 KB logs can be ingested at 200 logs/sec (2000 KB/sec).
- 1 KB logs can be ingested at 1000 logs/sec (1000 KB/sec).
This shows that the collector performs better when it is made to ingest bigger log entries (which is expected due to less overhead coming from timestamp parsing, etc.).
Memory usage guidelines
Benchmark - memory usage for particular average message size and EPS
Measured memory usage (in MB) for particular Events Per Second (EPS) average message size.
Fine Tuning Performance
There are a couple configuration options that can help with performance in specific scenarios.
Sumo Logic Exporter
The Sumo Logic Exporter sends data to Sumo Logic. It has the following features that can help with performance:
Read more about these features in the Sumo Logic Exporter docs.
The Batch Processor joins records of each type in batches. It has the following features that can help with performance:
Read more about these features in the Batch Processor docs.
Memory Limiter Processor
The Memory Limiter Processor prevents out-of-memory crashes for the collector process by monitoring the amount of memory used by the collector and forcing it to lower its memory consumption.
Read more about its features in the Memory Limiter Processor docs.