A Faraday cage keeps outside interference out. We study how well modern shared infrastructure does the same.
Almost every workload today runs on hardware it shares with strangers. Virtual machines compete for CPU, caches and network queues; containers share kernels; tenants share uplinks. The Faraday Laboratory measures how much one tenant can disturb another — and how much a tenant can tell about its neighbours.
Our work ranges from detecting noisy-neighbour effects in performance telemetry to verifying that network isolation policies actually hold in practice. We publish benchmarks and open tooling so that operators can check their own environments.
Lab lead: Dr. Elena Marchetti
Separating genuine application regressions from interference caused by co-located tenants, using CPU steal, cache and queueing signals.
Automated checks that network segmentation and firewall policies behave as declared, across providers and virtualisation stacks.
How much can be inferred about co-located workloads from timing alone — and which mitigations are cheap enough to deploy.
We welcome joint research, student projects and dataset requests. Write to lab@faraday.network-labs.sm-ventures.ru.