CURRENT_MEETING_REPORT_ Reported by Claudio Topolcic/BBN MINUTES The CO-IP Working Group met at the February 6-9 IETF Meeting at Florida State University. The Tuesday sessions were a presentation and discussion on ATM networks by Guru Parulkar of Washington University. The Wednesday morning session was a discussion of the issues, questions, and experiments raised by a guaranteed service network. The Wednesday afternoon session was canceled so that the working group members could attend the SMDS working group meeting. Work on the ST-2 protocols specification was dropped due to insufficient time. The ATM presentation consisted of roughly three parts: a BISDN perspective, ATM networks, and SMDS. The Broadband Integrated Services Digital Network is intended to support voice and video, DS1/3, X.25, and Switched Multi-megabit Data Services, with the latter including connectionless 802.6. The video services would include broadcast/permanent connections (e.g., cable TV), point-to-point connections, and switched connections, including conferencing. Video rates include both NTSC (45 Mbit) and ATV (135 Mbit). It is predicted that after 1995 it will be cheaper to run fiber to a home than to run copper. The protocol stack consists of application supported by an adaptation layer (which would include segmentation and reassembly if required) over the ATM layer over a SONET physical layer. SONET STS-3c consists of a 155.520 Mbit channel divided into 125 microsecond frames, with each frame containing 90 bytes of overhead and nine "channels" of 260 bytes each (the channels are not all byte aligned). An ATM frame consists of a 5-byte header followed by 48 bytes of data. The header format isn't yet standardized, but would most likely consist of 28 bits of combined Virtual Path Identifier (VPI) and Virtual Channel Identifier (VCI), an 8-bit checksum, a 2-bit priority field, and a 2-bit type field. A Virtual Path may inlude a number of Virtual Channels switched as a unit, so either the VPI or the VCI is used for cell forwarding on a hop-by-hop basis. The boundary between the VPI and VCI might vary at different interfaces. The VPI/VCI field might include other logical subfields, e.g., flow control information, etc. The Adaptation layer consists of a convergence sublayer on top of a segmentation and reassembly sublayer. The convergence sublayer wraps the padded application data with a header and trailer; the segmentation and reassembly sublayer segments the wrapped application data and adds its own header and trailer before passing each segment to the ATM layer. The services provided by ATM include point-to-point, multicast, and dynamic multicast callees, a QOS (which would probably be a fixed delay and loss specification within a homogeneous network), naked (aka dark) cells which will not be reordered by the network, and a bandwidth requirement specification. Bandwidth would be specified in terms of 1 mean, peak, and burst characteristics, with the actual nature of the latter still unspecified. Bandwidth consumption would be limited to that requested. With respect to CO-IP, there are two basic assumptions: the Internet will be heterogeneous for some time, and that LANs will not be ATM networks in the forseeable future. The conclusions were: since applications may generate packets larger than cell size, transparent fragmentation and reassembly should be supperted, CO-IP parameters should be consistent with ATM (at least in the voice context where the packet size is small) and CO-IP should try to be consistent with ATM naked cells (to minimize as much as possible the adaptation layer), the working group should make concrete plans for CO-IP experiments across ATM based high-speed networks, and to identify work that has/is being done in the ATM community for use in the CO-IP subnet dependent layer. Wednesday morning's session consisted of a discussion of CO-IP issues, questions, services and parameters. Included were: adherence to a schedule, blocking and delay, chokepoints, effect of linear topology problems and multi-hop paths, enforcement to meet performance requirements, fairness, reuse of unclaimed reserved bandwidth, combining best-effort and resource-reservation algorithms, throughput, and traffic characterizations. The latter were described as duration relative to RTT (i.e., << 2 RTT, 2 RTT, and >> 2 RTT), flowrate (steady, compressed steady, bursty), and predictability (none, e.g., interactive, ASAP, e.g., mail, and scheduled, e.g., a conference). A subset of the working group met Wednesday and Thursday evenings to discuss the practical details of future research collaboration. We agreed that such cooperation was possible, and would result in increased results with an overall decrease in effort. Since most participants felt most comfortable working with UNIX, we decided to adopt it as the experimental platform. We agreed to implement a basic protocol infrastructure in the UNIX-based DRI experimental gateway for experiments across the DRI testbed. MCHIP, ST-2, or other experimental protocols will be built on top of this infrastructure, and this would support experimentation and changes to the protocols. It will be possible to replace the modules that implement different functions, such as resource management or failure detection, relatively easily. By experimenting with them, we will gain practical experience in how different algorithms perform in various situations. These initial implementations will evolve to a single better protocol as we incorporate the better approaches. We are initially planning to implement a MCHIP gateway and host, and an ST-2 gateway and voice and video hosts. ATTENDEES Brim, Scott swb@devvax.tn.cornell.edu Casner, Steve casner@isi.edu Chatterjee, Samir samir@nynexst.com 2 Clapp, George meritec!clapp@bellcore.bellcore.com Easterday, Tom tom@nisca.ircc.ohio-state.edu Fidler, Mike ts0026@ohstvma.ircc.ohio-state.edu Fox, Richard sytek!rfox@sun.com Gerich, Elise epg@merit.edu Goldstein, Steve goldstein@note.nsf.gov Lynn, Charles clynn@bbn.com McKenney, Paul E. mckenney@sri.com Medin, Milo medin@nsipo.nasa.gov Parulkar, Guru guru@flora.wustl.edu Piscitello, David dave@sabre.bellcore.com Ramakrishnan, K.K. rama%erlang.dec.com@decwrl.dec.com Su, Zaw-Sing zsu@tsca.istc.sri.com Topolcic, Claudio topolcic@bbn.com Wilder, Rick rick@gateway.mitre.org Yavatkar, Raj raj@ms.uky.edu 3