The Airlock Anomaly Shield evaluation performance is limited by the number of parallel machine learning service worker processes. By default, the number of worker processes is 2, but can be increased if your system configuration allows assigning more exclusive machine learning service processes. Note that assigning additional worker processes will occupy CPU cores.
- We recommend the following steps:
- Check the system performance with Airlock Anomaly Shield enabled. Calculate a metric that examines how many evaluations take place per session per minute.
- Increase the number of worker processes in the
/opt/airlock/ml-service/conf/ml.ini
file accordingly.