Speaker: Xiaosong Ma
Time: 10:00-11:00 a.m., July 26, 2023, GMT+8
Venue: Room 1131, Science Building #1 (Yanyuan)
Abstract:
DRAM has remained a crucial resource for cloud/data-center applications, critical to both performance and cost. This talk will introduce two of our recent explorations that approach the problem from different angles. The first one examines the role disaggregated persistent memory (PM) could play in cloud-native relational database executions and proposes PilotDB, a new design that aims at making use of all major PM features: capacity, density, speed, and persistence, and byte-addressability. Indirectly, PilotDB retains the performance of a DRAM-resident database while dramatically reducing the (local) DRAM requirement, enabling the use of leaner, more agile compute nodes for fast deployment and recovery. The second study investigates the causes of inefficiency in today’s common memory cache design and finds cache management, typically implemented through linked-list maintenance, often does more harm than good. As a solution, our propose FrozenHot, a cache redesign that replaces constant cache management with periodic cache content update. By optimizing the hit path, it allows applications with good locality – who benefit most from memory caches – to get much closer to actual DRAM speed.
Biography:
Xiaosong Ma is currently a Principal Scientist at Qatar Computing Research Institute. Her research interests are in graph systems, distributed/cloud computing, and storage systems. Xiaosong has published over 100 research papers and currently serves on the editorial board of the ACM Transactions on Storage. She received both the DOE Early Career Principal Investigator Award and the NSF CAREER Award, and was named a University of Illinois Department of Computer Science Alumni Distinguished Educator and an ACM Distinguished Member. Xiaosong received her Ph.D. from the University of Illinois at Urbana-Champaign in 2003 and her B.S. from Peking University in 1997.
Source: School of Computer Sciences
