REIJS@SURFNET.NL (05/23/91)
1 Introduction Charing bandwidth is one of the main principles of network protocols. This can e.g. be done by means of time multiplexing (TDM) equipment and by means of packet switching (e.g. X.25, CLNP or DoD-IP). Within the framework of ISDN, CCITT has defined a level two protocol named Frame Relay, which is a light weight packet switching protocol. It has the connection oriented features (PVC's, accounting) of X.25, but it has e.g. high throughput, low end-to-end delay and no error correction. Frame Relay is part of CCITT I.122, but up until now it is still not a stable recommendation. To speed the implementation, some 20 suppliers made an agreement to make products, which will follow a defacto standard [1]. Both network provision and CPE (Customer Premises Equipment) implementations will be produced (see chapter 4). Because of the above characteristics, Frame Relay is a possible protocol to be used within a European backbone. The following sections will elaborate the possibilities of using Frame Relay in such a backbone. 2 Technology Frame Relay is belonging in the family of packet switching protocols. And like X.25 or DoD-IP it is also based on frames with a variable length (these protocols are called frame relaying technics). 2.1 The protocol It is a DTE-DCE interface level 2 protocol with a V.35 physical interface. Up until now it is only defined for T1/E1 (1.5/2 Mbps), but in future T3/E3 (45/34 Mbps) will be included in the standard. The defacto standard [1] is only based on permanent logical links (comparable with PVC's in X.25, Permanent Virtual Circuits). Within CCITT one is busy expanding the standard to switched logical links (Frame Relay 2). Like in X.25, a permanent logical link gets on both sides of the connection a (different) number, which are called DLCI's (Data Link Connection Identifier). Just like the LCN (logical Channel Number) in X.25, the DLCI has only local significance. Figure 1 gives an impression of the frame format. 2.2 The service Frame Relay is a level two service, it includes the following features: - variable frame length (the standard talks about a minimum allowable frame length of 262 octets, but at this moment the smallest implemented maximum frame length is 4000 octets). - maximum number of permanent logical links is 910 - no error recovery - frames arrive in the correct order - predefined permanent logical links are addressable by means of DLCI - low end-to-end delay (appr. 2 msec per node) - high throughput (because of light weight protocol) - no congestion control (must be preformed by DTE). Optional services (so not mandatory): - global addressing - multicast capability (for the connection less world this options will be very important for Address Resolution Protocols [ARP]). - flow control 2.3 Weaknesses of protocol Besides the advantages of Frame Relay, it also has some disadvantages. The following are the most important: - the flow control must be handled by the DTE's. If these are not able to do something, packets are lost in the Frame Relay network (level 3 or 4 protocols will handle this loss). - it is not clearly known how a Frame Relay network will handle congestion. And related to this: priority, use of discard facilities, rerouting - it is not yet know what performance information will be available from the Frame Relay network. 2.4 Management facilities Additional to the data link service, Frame Relay also has some management functions incorporated. These features are not yet incorporate in the CCITT standard. The defacto standard has the following properties: - the implementation of the management protocol is analog to Q.931 - it is DTE-DCE protocol - the DTE gets notifications on the change of DLCI's - the network is able to monitor the link between the DTE and DCE. An optional LMI-feature is asynchronous DLCI status updates. 3 The status of the standards 3.1 Frame Relay The formal standardization is being undertaken by CCITT and ANSI. CCITT defines the basic Frame Relay service in I.122. It defines two services: - Frame Relay 1 permanent logical links - Frame Relay 2 switched logical links The service is using the LAPD multiplexer method (Q.921) and it uses the out of band signaling of Q.931 (including management functions [LMI]). The defacto standard [1] is based on Frame Relay 1 and the LMI message format is based on Q.931. All mentioned standards are still not stable. The expected arrival of a first draft of the Frame Relay 2 will be end 1991. Some remarks on naming of standard (Q.xxx and I.xxx belong to CCITT and T1.xxx belong to ANSI): I.441 is identical to Q.921 T1.602 is equivalent to Q.921 I.451 is identical to Q.931 T1.607 and T1.608 are equivalent to Q.931 T1.606 is equivalent to I.122 3.2 Layers below Frame Relay At this moment suppliers have there own proprietary protocols below Frame Relay. Most protocols are cell relaying technics (frames with fixed lengths). In future Frame Relay will be transported by cell relaying protocols like ATM or DQDB. 3.3 Levels above Frame Relay There are not yet stable standards for supporting X.25, CLNS or DoD-IP over Frame Relay. In the Internet world people are busy making a RFC for DoD-IP over Frame Relay. The ongoing discussion can be followed through the list: IPLPDN@nri.reston.va.us (IP over Large Public Data Networks). No firm dates can be given when these protocols are stable. 4 Product availability Some 22 supplier did sign an agreement which said that they would implement Frame Relay (based on [1]). The suppliers can be divided in CPE and network equipment suppliers. De speed of the interfaces range from 256 kbps to 2 Mbps. It is expected that at the end of this year the products will be able to support 2 Mbps. The dates of the availability of products has been gathered in the period of Jan. up and including April 1991. In case not enough information is gotten, one sees question marks. 4.1 Frame Relay networks The following suppliers agreed on making Frame Relay network equipment (between brackets: date available, product name): AT&T (now, IACS), Hughes Networks (Q2/91, IPN), Newbridge (???, 3600/3645), Northern Telecom (Q4/91, DPN100), StrataCom (now, IPX), Timeplex (Q4/91, LINK), US Sprint (now, TP7900). 4.2 Levels below Frame Relay Only proprietary cell relaying technics exists (e.g. StrataCom). 4.3 Frame Relay CPE The following suppliers have agreed on making Frame Relay CPE: ACC (???, ACS4100), Cisco (now, AGS), Digital (???, ???), IBM (???, 3745), Netrix (???, ???), Timeplex (???, TIME/LAN), VitaLink (???, ???) and Wellfleet (???, ???). 5 Conclusions The overall conclusions are: - Frame Relay is still in the process of standardization - It is not yet feasible to make a European wide multiprotocol backbone, based on Frame Relay. If in the very near future products will be stable and tested, it could be used for a production multiprotocol backbone. - Frame Relay offers a service which could be very well used in multiprotocol backbone, because it could deliver a common protocol for X.25, CLNP and DoD-IP. 5.1 Future study subjects for multiprotocol backbone based on Frame Relay The following aspects must be studied: - What impact does the lack of congestion control have on an operational service? What additional features must be available within the Frame Relay protocol. - How important is it to have stable standards for routing/bridging X.25, CLNP or DoD-IP over Frame Relay? - Because Frame Relay is only an interface standard, there are not yet interworking standards between Frame Relay networks. What problems will result from this lack? - Why are there no European suppliers for Frame Relay equipment? 6 Documents [1] Frame Relay specification with extensions, doc. nr. 001-208966, 1990, Digital, Northern Telecom, StrataCom. [2] Broadband networks, McQuillan, J.M., Data communications international, June 1990, pp 62-74 [3] Frame Relay white paper, Button, B., StrataCom, May 1990. [4] The Frame Relay solution, US Sprint, Telenotes Vol. 1, Nr. 2, Sept. 1990. [5] Frame Relay networks: not as simple as they seem, Opderbeck, H., Data communications international, Dec. 1990, pp 89-91. [6] Review of backbone technologies, Heinanen, J., RARE High speed symposium, Jan. 17, 1991. [7] I.122, Framework for providing additional packet mode bearer services, Melbourne, Nov. 1988. [8] Q.921, ISDN user-network interface - Data link layer specification, Melbourne, Nov. 1988. [9] Q.931, ISDN user-network interface - Layer 3 specification for basic call control, Melborne, Nov. 1988.