Check out the new USENIX Web site.

Efficient Packet Demultiplexing for Multiple Endpoints and Large Messages

 
   Masanobu Yuhara                 Brian N. Bershad
Fujitsu Laboratories Ltd.       Department of Computer Science
1015 Kamikodanaka               and Engineering FR-35
Nakahara-ku                     University of Washington
Kawasaki  211, Japan            Seattle, WA  98195
 
   Chris Maeda                       J. Eliot B. Moss
School of Computer Science        Dept. of Computer Science
Carnegie Mellon University        University of Massachusetts
5000 Forbes Ave.                  Amherst, MA  01003
Pittsburgh, PA  15213

Abstract

(This research was sponsored in part by The Advanced Research Projects Agency, Information Science and Technology Office, under the title ``Research on Parallel Computing'', ARPA Order No. 7330, issued by DARPA/CMO under Contract MDA972-90-C-0035, by the Advanced Research Projects Agency, CSTO, under the title ``The Fox Project: Advanced Development of Systems Software'', ARPA Order No. 8313, issued by ESD/AVS under Contract No. F 19628-91-C-0168, by Fujitsu Laboratories Ltd., the Xerox Corporation, and Digital Equipment Corporation. Bershad was partially supported by a National Science Foundation Presidential Young Investigator Award. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Advanced Research Projects Agency, Fujitsu Laboratories, Xerox Corporation, Digital Equipment Corporation, the National Science Foundation, or the U.S. Government. Bershad performed this work while at Carnegie Mellon University. Authors' email addresses are yuhara,bershad,cmaeda,moss @cs.cmu.edu.}

This paper describes a new packet filter mechanism that efficiently dispatches incoming network packets to one of multiple endpoints, for example address spaces. Earlier packet filter systems iteratively applied each installed filter against every incoming packet, resulting in high processing overhead whenever multiple filters existed. Our new packet filter provides an associative match function that enables similar but not identical filters to be combined together into a single filter. The filter mechanism, which we call the Mach Packet Filter (MPF), has been implemented for the Mach 3.0 operating system and is being used to support endpoint-based protocol processing, whereby each address space implements its own suite of network protocols. With large numbers of registered endpoints, MPF outperforms the earlier BSD Packet Filter (BPF) by over a factor of four. MPF also allows a filter program to dispatch fragmented packets, which was quite difficult with previous filter mechanisms.

Download the full text of this paper in ASCII (43,750 bytes) form.

To Become a USENIX Member, please see our Membership Information.