fos - Building a Self-Aware Operating System for Multicores and Clouds
Clouds and multicores are two disruptive technologies that are altering the computing landscape dramatically by
offering virtually unbounded amounts of computational power in a cost effective manner. Managing all of these resources in a scalable manner in the face of elastic application demand, energy constraints and resiliency requirements is a huge challenge. Existing operating systems fall short as they are oblivious to energy constraints,
do not scale to large numbers of cores, and do not support clouds.
This talk will describe the Factored OS (fos) project which is creating a new OS to address the challenges of multicores
and clouds (http://groups.csail.mit.edu/carbon). To enable scalability, fos divides OS functions into separate system services, where each service is further factored into a parallel fleet of cooperating servers that communicate via message passing. The talk will also discuss how fos uses self-aware computing techniques such as Heartbeats and feedback control systems to manage the oft-conflicting goals of performance and energy efficiency even as application demand varies.
Anant Agarwal holds a Ph.D. (1987) and an MS in Electrical Engineering from Stanford University. He got his bachelor's
degree in Electrical Engineering from IIT Madras (1982). He is currently a professor of Electrical Engineering and
Computer Science at MIT, where his teaching and research interests include VLSI, computer architecture, compilation,
and software systems. He led a group that developed the MIT Alewife, a scalable shared-memory multiprocessor.
He also led the VirtualWires project at MIT and was founder of Virtual Machine Works, Inc., which took the
VirtualWires logic emulation technology to market. He currently leads the Raw project in the Computer Science and Artificial Intelligence Laboratory. He is also founder of InCert Software Corporation, which is applying a software agent technology to enterprise applications management.