DATE: Monday, August 19, 2013
TIME: Noon - 1:00 pm
PLACE: CIC 4th Floor ISTC Panther Hollow Conference Room
SPEAKERS: Jinyang Li, New York University
TITLE: Re-designing Distributed Systems for the Era of Fast Data-center Networks
ABSTRACT:
Existing datacenter-scale distributed systems are designed with the limitations of Ethernet in mind. Recent technological trends indicate that future datacenter networks will likely incorporate High Performance Computing network features, such as ultra-low latency and CPU bypassing. How should we design distributed systems differently to best utilize these features?
Will the re-design bring significant performance advantages? I will describe an ongoing project at NYU that aims to answer these questions. We have recently built a distributed in-memory key-value store called Pilaf that takes advantage of Remote Direct Memory Access, a form of CPU-bypassing, to achieve high performance with low CPU overhead. Specifically, Pilaf can surpass 1.3 million ops/sec using a single CPU core compared with 55K for Memcached and 59K for Redis.
BIO:
Jinyang Li is an associate professor of computer science at New York University. Her research interests include distributed systems and networks.
Her recent projects focus on developing better programming and storage infrastructure to simplify the scaling of applications in the cloud. Her honors include a NSF CAREER award (2007) and a Sloan Research Fellowship (2010). She received her B.S. from National University of Singapore and her Ph.D. from MIT, both in Computer Science.
HOST: David Andersen
VISITOR COORDINATOR: Kathy McNiff, 8-5099
SDI / ISTC SEMINAR QUESTIONS?
Karen Lindenfelser, 86716, or visit www.pdl.cmu.edu/SDI/