Seminar: Efficient Federated Learning Storage for non-training workloads
Ali Butt
Professor
Department of Computer Science and
Electrical Engineering
Virginia Tech
Friday, November 7, 2025
2:30 - 3:45 p.m.
Classroom Building, Room 260
Abstract
Federated Learning (FL) is an approach for privacy-preserving Machine Learning (ML), enabling model training across multiple clients without centralized data collection. We observed that a substantial part of FL systems are the non-training workloads such as scheduling, personalization, clustering, debugging, and incentivization. Extant systems rely on a central aggregator to handle non-training workloads and use cloud services for data storage. This results in high latency and increased costs as non-training workloads rely on large
volumes of metadata, including weight parameters from client updates, hyperparameters, and aggregated updates across rounds, making the situation even worse. I will present the design of FLStore, a serverless framework for efficient FL non-training workloads and storage, which addresses these challenges. FLStore unifies the data and compute planes on a serverless cache, enabling locality-aware execution via tailored caching policies to reduce latency and costs. Our evaluation shows significant improvements over the state-of-the-art, and our design fits easily with existing systems for easy adaption in practice.
Biography
Ali is a professor of computer science and ece at virginia tech. His research is on designing scalable I/O and storage systems for resource-intensive applications such as emerging generative AI and high-performance computing scientific workflows.