Published in IEEE INFOCOM, 2023
This paper addresses the challenge of optimizing cost and performance in deep learning clusters with separated compute and storage by proposing strategies to alleviate IO bottlenecks through caching or bandwidth scaling, tailored to the heterogeneous needs of different DL models and job characteristics.
Published in EuroSys, 2023
This paper proposes a deep learning scheduling framework that optimizes resource allocation by considering cache and remote IO, leading to substantial improvements in job completion time, utilization, and fairness.
Published in IEEE/ACM Transactions on Networking, 2021
This paper introduces two efficient algorithms for online file caching in multiple caches, outperforming current methods by reducing costs and maintaining high performance metrics.
Published in Knowledge-Based Systems, 2018
This paper is about density-based image clustering.