Skip to main content

IBM Israel Research Seminars

 

This talk is about software management systems for clusters and grids. Such systems consist of algorithms, that use system-wide feedback to optimize some operations e.g. cache management; and mechanisms which respond to explicit user's requests, without using such feedback. Traditionally, single node (including SMPs) management is done by the operating system (and the hardware), which in turn consist mostly of mechanisms.

The first part of the talk reviews existing cluster management packages, the majority of which consists of mechanisms. Most of these packages are suitable for high performance computing but inadequate for other applications.

The next part reviews the MOSIX cluster management approach, including its management algorithms to create a single system image, its method to support massive parallel I/O and its perfor-mance.

The last part of the talk shows how the MOSIX approach could be generalized to create a grid-wide virtual services, e.g., for security, load-balancing and for efficient management of grid resources.

About the speaker

Amnon Barak is Strauss Professor of Computer Science and director of the distributed systems research lab in the department of Com-puter Science at The Hebrew University of Jerusalem.

He received the B.S. degree in mathematics from the Technion, and the M.S. and Ph.D. degrees in computer science from the University of Illinois at Urbana-Champaign.

His primary research interests include parallel and distributed systems, management systems for clusters and grids, dynamic resource allocation, parallel I/O, high availability and competi-tive algorithms for resource management in clusters and grids.