Skip to content. Skip to main navigation.

From Resource Disaggregation to Cooperative Memory Expansion in Networked Computing Systems

Wednesday, October 18, 2017, 11:30 AM
ERB 228

Add to Calendar

Naan-Feng Tzeng, Ph.D.
University of Louisiana at Lafayette

Abstract:

Networked computing systems usually consist of commodity servers, which are configured with specific amounts of resources in CPU cores, DRAM, and storage. When executing diverse applications with varying resource requirements, such a system is often hard to best meet the resource needs of a given application dynamically during execution, making hardware resources hosted in each server being oversubscribed or overprovisioned. To this end, resource disaggregation has been proposed to address imbalanced resource requirements and to promise flexible scaling of individual resource types independently on-demand. Typical resource disaggregation establishes separate pools of computing resources, but it calls for extensive resource type-specific hardware and software knowledge. We have pursued resource disaggregation in an efficient and lightweight fashion, focusing on cooperative memory expansion (COMEX) in networked computing systems. COMEX establishes immense physical memory collectively across networked nodes on-demand, drastically accelerating application runs. Its functionality is seamlessly integrated with the Linux page-frame reclaiming function to exploit kernel information for superior page-in/out support while avoiding excessive overhead. A testbed with 12 servers networked by RDMA-enabled switch gear and with COMEX support has been deployed for evaluation. Evaluation results under ten benchmarks from two suites reveal that COMEX can achieve speedups exceeding 97× over its native OS counterpart.

Bio:

Nian-Feng Tzeng is Lockheed Martin Professor with the Center for Advanced Computer Studies, School of Computing and Informatics, University of Louisiana at Lafayette, which he joined in 1987. He has served as Associate Director of the School of Computing and Informatics since 2016. His current research interest is in the areas of high-performance computer systems, computer communications and networks, dependable computing and networked systems. He served on the editorial board of the IEEE Transactions on Parallel and Distributed Systems from 1998-2001, and on the editorial board of the IEEE Transactions on Computers from 1994-98. He was the Chair of Technical Committee on Distributed Processing, the IEEE Computer Society from 1999-2001. An IEEE Fellow, Tzeng is the recipient of the outstanding paper award at 10th International Conference on Distributed Computing Systems, May 1990. He received the University Foundation Distinguished Professor Award in 1997.

Return to Upcoming Events

 ∧