Multi- and Meta-Model Simulation and Analysis for datacenter infrastructure Collaborative Datacenter Simulation and Exploration for Everybody
Datacenters are vital for the digital society but represent a considerable fraction of global energy consumption. To improve their sustainability and performance when demand is foreseen to increase, we envision simulators and simulation-based digital twins will become primary decision-making tools. However, unlike other fields focusing on key societal infrastructure such as waterworks and mass transit, datacenter simulators cannot yet combine multiple, independent models into their operation. Addressing this challenge, in this work we propose M3SA, a datacenter simulation and analysis framework that uses discrete-event simulation to predict, per model and combined into a meta-model, the impact on climate and performance of various realistic datacenter conditions. We design an architecture for simulating with multiple concurrent models, a technique to integrate the results of multiple models into a meta-model, and a procedure to evaluate the accuracy of the meta-model.
M3SA is distributed under the MIT license. See LICENSE.txt.