Achieving speed and flexibility by separating management from protection: embracing the Exokernel operating system

ABSTRACT

The modern operating system is currently caught in a tug-of-war between two forces. At one end, there is a force that is demanding that the operating system become more flexible to handle the needs of evolving hardware and evolving user applications. At the other end, there is a force that is demanding that the operating system become more efficient to meet the needs of faster hardware. If the modern operating system does not keep pace with these two forces, it could cause the progress in computer design to become stagnant.

One possible solution to this problem is the Exokernel Operating System - an extensible (or easily modifiable) operating system developed at the Massachusetts Institute of Technology. Extensibility allows the operating system to be flexible to change and also open to optimization.

Extensibility within an operating system has resulted in several new issues. For example, extensibility seems to make customer-support harder to provide. Furthermore, some multithreaded applications perform worse in an extensible environment. Lastly, some have argued that it is optimization, not extensibility, that should be credited for the enhanced operating system speeds.

In this paper, we discuss the Exokernel Operating System with some detail. We explore some of the issues that have kept the Exokernel design from being the main-stream approach. We propose solutions to these issues and we conclude by trying to motivate the reader to embrace the Exokernel approach.

REFERENCES

• 1: "Practice and Technique in Extensible Operating Systems" by Lee Carver, Ying-Hung Chen, and Theodore Reyes. University of California - San Diego, 1998.
• 2: "Multiprocessing with the Exokernel Operating System" by Benjie Chen. Massachusetts Institute of Technology, February 2000.
• 3: "Extensible Kernels are Leading the OS Research Astray" by Peter Druschel, Vivek Pai, and Willy Zwaenepoel. Rice University, 1997.
• 4: "The Exokernel Operating System Architecture" by Dawson R. Engler. Ph.D. thesis, Massachusetts Institute of Technology, October 1998.
• 5: D. R. Engler , M. F. Kaashoek , J. O'Toole, Jr., Exokernel: an operating system architecture for application-level resource management, Proceedings of the fifteenth ACM symposium on Operating systems principles, p.251-266, December 03-06, 1995, Copper Mountain, Colorado, United States
• 6: Gregory R. Ganger , Dawson R. Engler , M. Frans Kaashoek , Hector M. Briceño , Russell Hunt , Thomas Pinckney, Fast and flexible application-level networking on exokernel systems, ACM Transactions on Computer Systems (TOCS), v.20 n.1, p.49-83, February 2002 [doi>10.1145/505452.505455]
• 7: "Exodisk: Maximizing application control over storage management" by Robert Grimm. Massachusetts Institute of Technology, May 1996.
• 8: Dejan Milojicic, Operating Systems-Now and in the Future, IEEE Concurrency, v.7 n.1, p.12-21, January 1999 [doi>10.1109/MCC.1999.749132]
• 9: "Why Aren't Operating Systems Getting Faster As Fast As Hardware?" by John Ousterhout. Digital Equipment Corporation, October 1989.
• 10: "User-Level Scheduling with Kernel Threads" by Thomas Riechmann and Jürgen Kleinöder. University of Erlangen-Nürnburg, June 1996.
• 11: "Shared Libraries in an Exokernel Operating System" by Douglas Wyatt. MS Thesis, Massachusetts Institute of Technology; August 1997.