OpenDC-STEAM: Realistic Modeling and Systematic Exploration of Composable Techniques for Sustainable Datacenters

A research team from VU Amsterdam has launched OpenDC-STEAM, an open-source, customizable simulation framework designed to tackle a critical gap in sustainable computing. Currently, techniques for reducing datacenter carbon footprints—like scaling resources, using on-site batteries, or shifting workloads to greener times—are often studied in isolation with simplified models. STEAM provides a realistic, holistic environment where these "composable techniques" can be tested together, accounting for real-world dynamics, diverse workloads, and fluctuating grid carbon intensity. This allows for a quantitative analysis of their combined effectiveness and the complex trade-offs they introduce between operational emissions, embodied carbon (from hardware), cost, and performance.

In their foundational analysis using STEAM, the team systematically explored three representative strategies: horizontal scaling (adding/removing servers), leveraging batteries for energy buffering, and temporal shifting of compute tasks. The simulator revealed that datacenter-specific dynamics significantly influence a strategy's success and that combining methods can substantially lower emissions, but creates a multidimensional optimization problem. By being open-source and extensible, STEAM aims to become a standard foundation for reproducible research, enabling operators to model their unique setups and test new sustainability methods before costly real-world deployment. The work, an extended version of a CCGRID 2026 paper, is available as free and open-source software (FOSS).