High-Performance Data Storage at Scale: Managing Petabytes of Data with Commodity Components

Scott A. Brandt, University of California, Santa Cruz

Tuesday, March 21
12:15pm - SENSQ 5317

Free pizza will be provided to attendees, starting at noon

Hosted by Daniel Mosse'

Abstract

High-performance computing places extreme demands on the storage subsystem. Future high-performance storage systems will have tens of thousands of clients, tens of petabytes of storage, files ranging from bytes to terabytes, and aggregate data rates of 1 terabyte/ second or more. Many well-understood issues become quite difficult at this scale, including reliability, data distribution, metadata management, and data storage. This talk discusses Ceph, our high- performance distributed object-based storage system targeted for such environments. We will discuss the unique challenges presented by high- performance data storage at scale, explain how Ceph addresses these challenges, and explore some remaining challenges and ongoing research in this area.

Biography of Speaker

Scott Brandt is Associate Professor of Computer Science at the University of California, Santa Cruz. He is co-founder and Associate Director of the UCSC Storage Systems Research Center and the UCSC/ LANL Institute for Scalable Scientific Data Management. Prof. Brandt received his Ph.D. in Computer Science from the University of Colorado, Boulder in 1999, and an M.S. in Computer Science in 1993 and a Bachelor of Mathematics in 1987, both from the University of Minnesota. Before joining academia, Prof. Brandt spent 10 years in industry doing research and development in real-time image processing systems, secure operating systems, and asynchronous circuit technology. Prof. Brandt's current research focuses primarily on high- performance storage systems and integrated and adaptive real-time systems.