EnOSlib multi-provider

EnOSlib is the experiment library the STACK and Discovery teams use to describe, deploy, and evaluate distributed systems on real infrastructures. Historically, an experiment targeted a single provider: you reserved machines on Grid’5000, or on a cloud, or on IoT-lab, but not all of them at once.

Together with Baptiste Jonglez, Matthieu Simonin, and Jolan Philippe, I contributed to extending EnOSlib with multi-provider capabilities, so that a single experiment description can span the whole edge-to-cloud continuum.

🧪 What the extension brings

• Synchronized reservations across several testbeds, so that heterogeneous resources are available in the same time window
• A uniform abstraction over providers with very different reservation models (batch schedulers, cloud APIs, sensor testbeds)
• Reproducible deployments: the same description drives bare-metal nodes, virtual machines, and constrained edge devices
• Support for realistic edge-to-cloud experiment topologies, including network emulation between the tiers

🎯 Why it matters

Experimental research on the continuum is only credible if the experiments can be re-run. Being able to describe a multi-tier deployment once, and replay it across providers, removes a large part of the manual glue code that usually makes such experiments fragile and unrepeatable.

This work was published at DAIS 2025 (25th IFIP WG 6.1 International Conference on Distributed Applications and Interoperable Systems, part of DisCoTec 2025, Lille), and directly supports the experimental setups used in my own observability and orchestration work.