Fairness-Aware Job Scheduling for Multi-Job Federated Learning

TL;DR

This work tackles fairness in multi-server federated learning where several publishers concurrently recruit clients from a shared pool. It introduces FairFedJS, a Lyapunov-optimization-based scheduler that uses virtual queues, a per-round Job Scheduling Index, and data-fairness-aware client selection to balance dataset demand and bid payments, while leveraging reputation from a Beta Reputation System. Payments follow a Derivative Follower Pricing rule, and the framework optimizes a drift-minus-utility objective to trade off scheduling fairness and system revenue, under a horizontal FL assumption with a single data type per job. Empirical evaluations on Fashion-MNIST and CIFAR-10 show substantial gains in scheduling fairness (about 31.9% improvement) and faster convergence (about 1.0% faster) with accuracy comparable to baselines. The approach is positioned as the first fairness-aware job scheduling method for multi-job FL, with planned extensions to vertical FL and federated transfer learning where jobs may require multiple data types.

Abstract

Federated learning (FL) enables multiple data owners (a.k.a. FL clients) to collaboratively train machine learning models without disclosing sensitive private data. Existing FL research mostly focuses on the monopoly scenario in which a single FL server selects a subset of FL clients to update their local models in each round of training. In practice, there can be multiple FL servers simultaneously trying to select clients from the same pool. In this paper, we propose a first-of-its-kind Fairness-aware Federated Job Scheduling (FairFedJS) approach to bridge this gap. Based on Lyapunov optimization, it ensures fair allocation of high-demand FL client datasets to FL jobs in need of them, by jointly considering the current demand and the job payment bids, in order to prevent prolonged waiting. Extensive experiments comparing FairFedJS against four state-of-the-art approaches on two datasets demonstrate its significant advantages. It outperforms the best baseline by 31.9% and 1.0% on average in terms of scheduling fairness and convergence time, respectively, while achieving comparable test accuracy.