Abstract (EN)

Motivated by today’s decentralized operation of interconnected computing platforms, classical task scheduling models are revisited under a game theoretic perspective. Instead of being designed by a central entity which aims at optimizing an aggregate efficiency measure, task allocations emerge through aggregated localized decisions taken by a group of autonomous self-interested agents. The outcome is sought as an equilibrium whose overall social efficiency typically diverges from the optimal group’s choice. This divergence, captured by a measure that came to be known as the Price of Anarchy, can be alleviated by local scheduling policies called Coordination Mechanisms. This chapter reviews standard task scheduling models, dedicated coordination mechanisms and their influence on the price of anarchy. It also exemplifies the design and analysis of coordination mechanisms on a particular scheduling model with setup times, and discusses open research questions.