Precomputed Dominant Resource Fairness

2025-07-08

AI Summary

This paper addresses fair allocation of multiple resource types in distributed systems. We propose an approximate Dominant Resource Fairness (DRF) algorithm leveraging a precomputation mechanism—the first to integrate precomputation into the DRF framework. By offline analyzing users’ dominant resource shares and preconstructing allocation policies, the algorithm drastically reduces online iteration steps. It rigorously preserves DRF’s core axiomatic properties: sharing incentive, envy-freeness, and strategy-proofness, while theoretically generalizing max-min fairness to multi-resource settings. Experimental evaluation in large-scale cloud and datacenter environments demonstrates that the algorithm achieves fairness comparable to exact DRF using only 30%–50% of the computational steps, thereby attaining a favorable trade-off between efficiency and fairness.

Technology Category

• Game Theory and Economic Paradigms: Fair Division
• Multiagent Systems: Mechanism Design
• Search and Optimization: Distributed Search

Application Category

• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Although resource allocation is a well studied problem in computer science, until the prevalence of distributed systems, such as computing clouds and data centres, the question had been addressed predominantly for single resource type scenarios. At the beginning of the last decade, with the introuction of Dominant Resource Fairness, the studies of the resource allocation problem has finally extended to the multiple resource type scenarios. Dominant Resource Fairness is a solution, addressing the problem of fair allocation of multiple resource types, among users with heterogeneous demands. Based on Max-min Fairness, which is a well established algorithm in the literature for allocating resources in the single resource type scenarios, Dominant Resource Fairness generalises the scheme to the multiple resource case. It has a number of desirable properties that makes it preferable over alternatives, such as Sharing Incentive, Envy-Freeness, Pareto Efficiency, and Strategy Proofness, and as such, it is widely adopted in distributed systems. In the present study, we revisit the original study, and analyse the structure of the algorithm in closer view, to come up with an alternative algorithm, which approximates the Dominant Resource Fairness allocation in fewer steps. We name the new algorithm Precomputed Dominant Resource Fairness, after its main working principle.

Problem

• Fair allocation of multiple resource types
• Extending Max-min Fairness to multiple resources
• Approximating Dominant Resource Fairness efficiently

Innovation

• Generalizes Max-min Fairness to multiple resources
• Approximates Dominant Resource Fairness efficiently
• Precomputes allocations for faster performance