USENIX SREcon21: Automating performance tuning with machine learning

Akamas will be at SREcon 2021 online event on October 14th at 2:00 UTC with his CTO, Stefano Doni. He will present a new approach to performance tuning based on machine learning techniques that help SREs to both reduce the toil and automate the process.

This new approach allows SREs to define an optimization goal, such as maximize throughput, minimize latency or cloud costs, and constraints reflecting any defined SLOs, such as keeping availability above and response time below given levels. Driven by the goal, the ML-based optimization is then able to automatically identify the best values for the configuration parameters to optimize the target system. 

In this session, Stefano will show how this new approach allows SREs to achieve higher systems performance, efficiency and reliability with a lot less manual work thus allowing SREs to focus on higher value tasks, such as setting optimization objectives and finding new optimization opportunities across the systems.: