Off loading the protocol

David Farber (
Fri, 18-Mar-88 06:38:56 est

The following is abstracted from a note on our experiences with a line
of research aimed at eamining a radically different approach to "off
loading" designed for the very high speed networking era -- "Gigabit
networking". I would be happy to supply the full text and/or memnet
documents to any interested people.

             Some Thoughts on the Impact of Very High Speed Networking on
                                Processor Interfaces



            This note proposes a completely different view of computer
          networking, a view which derives from experiments started in my
          group at the University of Delaware (and continuing at the
          University of Pennsylvania) resulting in the creation a novel
          local network architecture called "MEMNet." MEMNet is a research
          system aimed at exploring ways of removing the severe processing
          overhead found in distributed operating systems. The approach
          MEMNet takes is to treat the network as a mechanism which allows
          a processor to access the collective memory space of the
          distributed system. Thus, when a processor in a MEMNet
          environment needs to send data via the high-speed local network,
          it simply writes to memory addresses which are in the memory
          space of the recipient processor. Similarly, the recipient
          processor, when it chooses to examine data which has been "sent"
          by another processor, reads its local memory (or physically-
          remote memory, in a hierarchical memory system) simply by the
          normal memory access mechanisms of that processor. In the MEMNet
          environment, there is a set of special memory controllers with
          adequate caching, connected together via a high-speed (200
          megabit) ring. The caching provides a mechanism equivalent to the
          snooping caches of modern multiprocessors.

            During this research, we examined the architecture of software
          systems which would run in a MEMNet environment. Much to our
          surprise (although we should not have been surprised), the
          software implications of such a distributed environment are
          essentially non-existent. That is, a software system written to
          run on the fully distributed MEMNet environment is essentially,
          in all respects, identical to the same software system designed
          to run on a simple multiprocessor, shared-memory environment.

            The issues one must face in designing the architecture of
          future wide-area, high-speed networks ......


            In summary, this note suggests that we reexamine what we mean
          by "networking" in the future. It essentially suggests that
          networking is simply a special case of interprocess communication
          over a widely-distributed computer system, and thus can take
          advantage of technology already developed.


David J. Farber; Prof. of CS and EE, Director - Distributed Systems Labs.
University of Pennsylvania/200 South 33rd Street/Philadelphia, PA 19104-6389
Tele: 215-898-9508; FAX: 215-274-8192

This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:41:06 GMT