Structuring operating systems using adaptive objects for improving performance

Abstract

Programmers of high-performance applications often complain that the abstractions provided by the operating system (OS) hinder the optimal exploitation of the underlying hardware. We believe that this should not be the case with a well designed OS. The key is that OS subsystems must be customizable, adaptable and extensible, to permit an optimal match between the application needs and the hardware architecture. In the Chameleon project, we seek to demonstrate the practical feasibility of dynamically customizing OSs in a distributed or clustered system. We intend to test our prototype with a small set of commercial applications.