Comments on Proposed Transport Protocols


Gurudatta Parulkar (guru@flora.wustl.edu)
Wed, 10 Aug 88 09:46:05 -0500


We just completed a DRAFT of a technical report with the abstract
enclosed. If you would like a copy to review, please send me a note.
I want to mention that this report does not contain any quantitative
analysis, and has only qualitative reasoning to support the claims.

-guru

Dr. Guru Parulkar
Asst Professor guru@flora.wustl.edu
Dept of Computer Science parulkar@udel.edu
Washington University wucs1!guru@uunet.uu.net
St. Louis MO 63130
(314) 889-4621
-------------------------------------------------------------------------

               Comments on Proposed Transport Protocols

                             Anil Bhatia
                            James Sterbenz
                            Guru Parulkar
                         guru@flora.wustl.edu

              Computer and Communication Research Center
                    Department of Computer Science
                        Washington University
                          St. Louis MO 63130

                               Abstract

Over the last few years, a number of research groups have made
considerable progress on the design of high speed networks -- on the
order of a few hundred Mbps to a few Gbps. The emphasis of this work
has been on the design of packet switches and on the design of network
access protocols. However, this work has not yet addressed the
internetworking and transport level issues in a high speed internet.

The ARPA Internet, which has contributed a number of fundamental ideas
to the field of internetworking of diverse networks, cannot be used in
its current form as a model for the very high speed internetworks
(VHSI) of the future.In this report, we propose a revised internet
model, called the VHSI model, which among other things, argues for a
number of application-oriented lightweight transport protocols
(ALTPs). As part of our effort on the design of a VHSI model, we
considered the appropriateness of recently proposed transport
protocols, NETBLT and VMTP, as ALTP candidates. The purpose of this
report is to present the results of this study, and present our
thoughts on the design of ALTPs.

The summary of the results of this study is that NETBLT and VMTP have
contributed a number of interesting ideas to the design of transport
protocols, and they do improve upon TCP within the current Internet
model for the applications they were originally designed for.
However, we believe that these protocols are not appropriate solutions
for the VHSI model, because the underlying assumptions and trade-offs
that these protocols are based on are very different in the VHSI
model. For example, the VHSI model assumes a {\em quasi-reliable}
connection-oriented internet protocol (as opposed to the current
unreliable datagram IP), which can make performance guarantees and can
ensure that the internet is congestion free (almost all the time).
Also, the network speeds in the VHSI are a few order of magnitude more
than what NETBLT and VMTP assume. We argue that the transport
protocols in the VHSI model should avoid end-to-end flow control as
much as possible, and make the end-to-end error control application
specific and indepedent of the end-to-end latency. In general, the
transport protocols should be simpler, designed to be mostly
implemented in VLSI, well integrated with the host architecture and
operating system, and targetted for a specific class of applications.

The rest of this report is organized as follows: Section 2 provides an
overview of the ARPA Internet model and the VHSI model. It also gives
an overview of TCP and its associated problems. Section 3 provides an
overview of NETBLT and VMTP and summarizes how these protocols are
improvements over TCP. Section 4 describes the limitations of NETBLT
and VMTP within the VHSI model. Section 5 presents our thoughts on the
design of ALTP protocols for VHSI, and finally, Section 6 is the
conclusion.



This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:43:12 GMT