TMux to Proposed Standard
Allison J Mankin <mankin@cmf.nrl.navy.mil> Fri, 12 August 1994 21:18 UTC
Received: from ietf.nri.reston.va.us by IETF.CNRI.Reston.VA.US id aa10932; 12 Aug 94 17:18 EDT
Received: from CNRI.RESTON.VA.US by IETF.CNRI.Reston.VA.US id aa10928; 12 Aug 94 17:18 EDT
Received: from basil.xylint.co.uk by CNRI.Reston.VA.US id aa17548; 12 Aug 94 17:18 EDT
Received: from radegond.cmf.nrl.navy.mil ([134.207.7.18]) by xylint.co.uk (5.0/SMI-SVR4) id AA03727; Fri, 12 Aug 1994 21:03:25 +0000
Received: from localhost (mankin@localhost) by radegond.cmf.nrl.navy.mil (8.6.8.1/8.6.6) with SMTP id QAA15133; Fri, 12 Aug 1994 16:02:45 -0400
Message-Id: <199408122002.QAA15133@radegond.cmf.nrl.navy.mil>
X-Authentication-Warning: radegond.cmf.nrl.navy.mil: Host localhost didn't use HELO protocol
To: barnes@xylogics.com, cameron@xylint.co.uk, postel@isi.edu, dcrocker@mordor.stanford.edu, cohen@rand.org
Cc: cmp-id@xylint.co.uk
Subject: TMux to Proposed Standard
Date: Fri, 12 Aug 1994 16:02:38 -0400
Sender: ietf-archive-request@IETF.CNRI.Reston.VA.US
From: Allison J Mankin <mankin@cmf.nrl.navy.mil>
content-length: 5761
Jim, Pete, Jon, Dave and Danny, Congratulations. TMux was passed as a Proposed Standard (with its Applicability Statement) by IESG vote yesterday. You will see the Protocol Action message before Monday. Best wishes, Allison Mankin Transport Area Director P.S. Preview of the Protocol Action writeup -- ----------------------- The IESG has approved the Internet-Draft "Transport Multiplexing Protocol (TMux)" <draft-cameron-tmux-03.txt> as a Proposed Standard. This document is not the product of an IETF working group, but has been reviewed in the Transport Area of the IETF. The IESG contact person is Allison Mankin. Technical Summary One of the problems with the use of terminal servers is the large number of small packets they can generate. Frequently, most of these packets are destined for only one or two hosts. The overhead (interrupt and other) for processing a short packet on a host or terminal server can be very high, and terminal server vendors have found that TCP/IP without a protocol like TMux performs poorly compared with some proprietary protocols such as Digital Equipment Corporation's LAT. TMux is a protocol which allows multiple short transport segments, independent of application type, to be combined robustly between a server and host pair. A TMux message appears as: | IP hdr | TM hdr | Tport segment | TM hdr | Tport segment| ...| Where: TM hdr is a TMux mini-header and specifies the following Tport segment. Tport segment refers to the entire transport segment, including transport headers. Each 4 octet TMux mini-header has the following general format: +-------------------------------+ | Length | | | | | +-------------------------------+ | Protocol ID | +-------------------------------+ | Checksum | +-------------------------------+ | Transport segment | | ... | | ... | The Protocol ID field contains the value that would normally have been placed in the IP header Protocol field. The 'Checksum' field is the XOR of the first 3 octets. TMux operates as an extension to the IP datagram protocol. Hence, it has no impact on most flow control mechanisms, since they operate at the transport layer -- above TMux. This was the subject of careful review by the Transport Area. The specification includes details of a dynamic and stateless method of turning TMux operation on and off, as well as a timeout for accumulating short segments with suggested value supported by performance measurements. TMux is proposed with an Applicability Statement because it is known by experiment to be effective in the case of a large terminal server in the local area, and it may not be applicable for other environments. Implementations are expected to not use TMux for packets of 700 bytes or more. A robust and fast mechanism is specified for turning the use of TMux on and off dynamically. The specification recommends also that routers use protocol filtering to not forward TMux traffic (it is IP protocol type 18). Hosts or servers that send a TMux ENQuiry message and do not receive TMux traffic or a TMUX ENQ in reply simply send unmultiplexed transport over IP datagrams. Performance results for this terminal server case were detailed in the TMux BOF at the Houston (November '93) IETF meeting. In its applicability environment, TMux is a valuable optimization. Working Group Summary At the Amsterdam IETF, Peter Cameron and Jim Barnes made an initial proposal to address the terminal server problem with CMP (connection multiplexing protocol), a layer over TCP. It was found to have architectural and performance faults, and the beginnings of the TMux proposal were offered by Dave Crocker. Jon Postel and Danny Cohen had made a very similar proposal to TMux in IEN 90 in YEAR. The strong consensus of those attending the Amsterdam CMP BOF was that the TMux approach was superior to CMP and highly promising. A TMux BOF was convened at the Houston IETF, to report on the refinement of TMux over email and on implementation experiences. Performance improvements fulfilling the goals of the proposal were reported. There was strong consensus in the Houston BOF that the protocol proposal was valid and should receive a few more corrections and then be submitted to the standards track. The Transport Area Director requested that an Applicability Statement be made with the protocol proposal, because of the specialized, though sizeable, constituency for a terminal server optimization. There were several rounds of revision of the protocol based on review in the Transport Area and the IETF in general. The Last Call on the Internet-Draft of the specification did not generate any responses, and there is consensus in the Transport Area Directorate that the proposal merits Proposed Standard status. Protocol Quality The protocol specification was extensively reviewed by the Transport Area Directorate, including written reviews by Dave Borman and Greg Minshall, and by the area director. Experimental implementations were done by two known groups and two vendors are known to have completed product-quality implementations.
- TMux to Proposed Standard Allison J Mankin
- Re: TMux to Proposed Standard Dave Crocker