Experimentation with a Reconfigurable Microkernel

Since the implementation of operating system functions can significantly affect the performance of parallel programs, it is important to customize operating system functionality for specific application programs. In this paper, we propose an architecture for a reconfigurable micro-kernel. This kernel can be configured at compile-time and at execution-time to suit varying application requirements. Such a micro-kernel can be used for the development of high performance operating systems and applications for parallel and distributed systems. We have implemented the reconfigurable micro-kernel on multiple parallel machines, including a 32-node GP1000 BBN Butterfly, SGI multiprocessors, and a 32-node Kendall Square Supercomputer.