Networks and Distributed Systems

Strymon: A system for predictive datacenter analytics and management via queryable online simulations

Modern enterprise datacenters are highly complex, multi-layered distributed systems whose components interact in often unpredictable ways. Consequently, it is hard to say in advance what will happen if the workload of the datacenter changes significantly, and making configuration changes, modifying policy, or deploying new services or capacity is often a leap in the dark.

We are building Strymon to address such problems. Strymon is a system for predicting the datacenter behavior in hypothetical scenarios based on the novel approach of queryable online simulation. Strymon leverages existing logging and monitoring pipelines of modern production datacenters to ingest cross-layer events in a streaming fashion and predict possible effects of such events in what-if scenarios. The actual predictions are made online by simulating the hypothetical datacenter state alongside the real one. More...


Back to Research