Vessim: A Testbed for Carbon-Aware Applications and Systems

TL;DR

Vessim addresses the lack of realistic testing environments for carbon-aware computing by offering a modular co-simulation testbed that links energy-system simulators with real computing workloads. Built on Mosaik, it integrates a microgrid-centric architecture with actors, a grid, controllers, and energy storage, enabling hardware-in-the-loop and software-in-the-loop experiments. The work contributes a Python API, integrated renewable and storage models, historical/live datasets, and an illustrative edge-computing scenario, all supporting reproducible evaluations and digital-twin style operation. It enables researchers to study grid balance via $Δp_t = ∑ p_t^a$ and energy-state transitions through $e_{t-1}$ under realistic energy dynamics, accelerating the development of carbon-aware strategies for distributed computing. The platform thus promises faster prototyping, better comparability, and practical impact for datacenters and edge deployments engaging with dynamic energy systems.

Abstract

To reduce the carbon footprint of computing and stabilize electricity grids, there is an increasing focus on approaches that align the power usage of IT infrastructure with the availability of clean energy. Unfortunately, research on energy-aware and carbon-aware applications, as well as the interfaces between computing and energy systems, remains complex due to the scarcity of available testing environments. To this day, almost all new approaches are evaluated on custom simulation testbeds, which leads to repeated development efforts and limited comparability of results. In this paper, we present Vessim, a co-simulation environment for testing applications and computing systems that interact with their energy systems. Our testbed connects domain-specific simulators for renewable power generation and energy storage, and enables users to implement interfaces to integrate real systems through software and hardware-in-the-loop simulation. Vessim offers an easy-to-use interface, is extendable to new simulators, and provides direct access to historical datasets. We aim to not only accelerate research in carbon-aware computing but also facilitate development and operation, as in continuous testing or digital twins. Vessim is publicly available: https://github.com/dos-group/vessim.

Figures (4)

• Figure 1: Vessim enables research on various aspects regarding interactions between computing and energy systems.
• Figure 2: Vessim allows the simulation of multiple Microgrids in parallel where each consists of the depicted components. Hexagons represent a co-simulation subsystem managed by Mosaik Mosaik_2019. Actors represent consumers/producers; the grid simulator determines the current excess/deficit in power; controllers can be used for monitoring, decision-making, or providing APIs to computing systems; energy storage simulators can charge/discharge batteries depending on user-defined policies.
• Figure 3: Scenario definition with Vessim’s Python API. The scenario connects a solar and battery simulation with a physical, power-metered device (NVIDIA Jetson TX2).
• Figure 4: Results of the example scenario. 1. $p^a_t$ of both actors, 2. $\Delta p_t$ computed by the grid simulator, 3. the battery's SoC (part of $state_t^s$), 4. energy deficit ($-e_t$) & MOER, and 5. the accumulated carbon emissions ($-e_t \cdot$ MOER).