A new communication protocol for accessing data networks

A new communication protocol for accessing data networks-The international packet-mode interface by A. RYBCZYNSKI BeU Canada Ottawa, Canada B. WESSLER Telenet Communications Corporation Washington, D.C. R. DESPRES Administration Francaise de PTT Rennes, France and J. WEDLAKE United Kingdom Post Office London, England manufacturers of data processing and terminal equipment and common carriers. ABSTRACT Public packet switching networks are at various stages of development around the world, notably in the U.S., Canada, France, the United Kingdom and Japan. The success of these networks is highly dependent on the use of an agreed-upon standard device-independent interface between the packet networks and the user devices operating in the packet-mode. This interface consists of far more than the data link control procedure (Le., HDLC), which administers the physical transmission medium between the data terminal equipment (DTE) and the network. The specification of the packet-mode interface defines a set of conventions governing the manner in which DTEs establish, maintain and clear calls, format control information and data into packets and manage the flow of data for many calls over a single circuit to and from the packet network. This paper describes the International Packet-Mode Interface, developed jointly by Telenet Communications Corp., the Trans-Canada Telephone System (TCTS), the United Kingdom Post Office and the French PTT. This interface has been designed to enable DTEs such as computers, programmable terminal controllers and intelligent terminals to gain access to public packet networks throughout the world. The present status of international standardization of this interface within the CCITT is also covered. Standardization of the International Packet-Mode Interface is to the advantage of teleprocessing users, INTRODUCTION Te!enet has been commercially available in the U. S. since August 1975; Datapac will start commercial operation in Canada next month; the French experimental RCP network started operation in 1975 and Transpac development has been contracted out and will start providing service in France in 1978; the Experimental Packet Switching Service of the United Kingdom Post Office is well into the evaluation stage; and other public data networks are being planned and developed in both Japan and the Nordic countries. The fundamental technology used in all of these networks is packet switching. In packet switching, all user data is formed into discrete units called packets. In addition to the data to be transferred (typically part of a message), each packet includes a header specifying control functions and the destination to which the data is to be delivered. Packets are routed through the network on a store and forward basis and travel very rapidly and accurately through the network, experiencing only a fraction of a second delay from source to destination. Additionally, the network performs buffering functions so that the speed and format of the data sent into the network can be different from those of the data received at the destination. The packet switching technology described briefly 477 From the collection of the Computer History Museum (www.computerhistory.org) 478 National Computer Conference, 1976 in the preceding paragraph is actually of little direct concern here, since the sophisticated routing, monitoring and error correction techniques internal to packet networks are invisible to users of the networks. (Readers are referred to a book edited by Chu l for a discussion of packet network design considerations). Rather, the user's primary interests are the characteristics of the service provided by a network and of the interface between him and the network. The public data networks listed above are supporting packet-mode services based on the virtual circuit concept. A virtual circuit is a bi-directional association between a pair of DTEs over which all data transfer takes the form of packets. Transmission facilities are only assigned when data or control packets are actually being transferred. The virtual circuit concept permits a DTE (data terminal equipment) to establish concurrent communications paths to many other DTEs over a single physical access circuit (Figure 1). The high degree of sharing made possible by the use of virtual circuits enables communication savings to be passed on to the user. The virtual circuit concept minimizes the impact of new packet switching services on existing user systems. Two types of interfaces may exist on any packet switching network. The first type may be called a device-dependent interface such as would be required at present for most hard-wired terminals (e.g., pointof-sale terminals, teletype machines). The second type may be called a device-independent interface which is applicable to most programmable devices (e.g., computers, programmable controllers, concentrators, intelligent terminals). Because of the large number of terminals that require device-dependent interfaces and because of the large incompatibilities that presently exist among these terminals, a large number of interface specifications are required. On the other hand, only a single specification of a device-independent interface is necessary. The International Telephone and Telegraph Consultative Committee (CCITT) in Geneva has recognized the need for these two types of interfaces and has put priority on the development of a recommendation for the second type of interface; that is, a deviceindependent interface between what it calls a packetterminal and the packet switching network. A similar conclusion has been drawn by an Ad Hoc Group of U. S. ANSI Task Group X3S37 (Public Data Networks) .2 Likewise, TCTS, the French PTT, Telenet, the Japanese NTT, and the United Kingdom Post Office have put high priority on the development and standardization of the International Packet-Mode Interface. The specification of the International PacketMode Interface was submitted as draft Recommendation X.25 3 to the CCITT Secretariat for consideration by the Study Group VII Plenary. The purpose of specifying the International PacketMode Interface is to provide an efficient means by which a large set of characteristically different teleprocessing systems can gain access to the services and related technical and economic benefits of packet switching networks. GENERAL DESCRIPTION Interface 1'equirements The basic requirements imposed upon the architecture of the interface are introduced below: , ~t-c=_."' ",;~'l ",," ,," I ,,",,"" ,," " ,,, ... -'" ... (DeE) Network I I ~ , - ... __ - ,Virtual '~ircult/ \ I \ I \ ~I I / / I ....... I """", , \ ' ' '~ \ Figure l-Use of virtual circuits 1. The interface shall provide a full duplex transmission path between the DTE and the network. 2. It shall ensure the integrity and accuracy of the data transmitted between the DTE and the network. 3. It shall provide the DTE with switched and permanent virtual circuits. 4. It shall be capable of efficiently supporting concurrent communication between a packet mode DTE and numerous other DTEs over a single physical circuit to the network. 5. It shall allow both DTE and network to control the flow of data over the access circuit so that one does not overload the other. 6. It shall provide supervisory and control functions to administer calls satisfactorily. 7. It shall do all of above using existing standards wherever possible. From the collection of the Computer History Museum (www.computerhistory.org) International Packet-Mode Interface Interface characteristics The International Packet-Mode Interface consists of three distinct levels of control procedures as illustrated in Figure 2: 1. the Physical Interface 2. the Frame Level Logical Interface 3. the Packet Level Logical Interface Each of these levels functions independently of the other levels, with the exception that failures at a level may affect the operation of higher levels. The Physical Interface specifies the use of a duplex, point-to-point synchronous circuit, thus providing a physical transmission path between the DTE and the Network. It also specifies the use of an existing physical interface (i.e., EIA RS-232-C standard) between the DTE and a data set or modem. Therefore, no changes to the interface hardware of the DTE are required. The Frame Level Logical Inter'face specifies the use I I I I I I I process i4---------------+------------------.... : To other User process I I I I -----1-------------n r-----· -----1--------..y____________ Packet Level procedure. I ~- ! I' ~-_ ______________ .. I I Frame Level II !4t-.~ I"\. procedures packet Level LOlical Interface (multi-channel) Frame Level LOllcel Interface (slnlle data link) .. Network Physical Level .i I' Proceduras Physical Interface (4-wlre· point-tapoint synchronous circult) 479 of a data link control procedure which is compatible with the High-Level Data Link Control (HDLC) procedures being standardized by ISO and with the Advanced Data Communications Control Procedure (ADCCP) being standardized by U. S. ANSI. The Frame Level Interface uses the principles of a new ISO Class of Procedure for a point-to-point balanced system, whereby the DTE and the network node each have a primary and a secondary function. The Frame T.aual ..L...oIv v '-".I. T,.,ta ...-f<:lt>a ;'" rlafl,.,arl ;,., ta ....I..I.J....:J 'rVIC' n-f n"';'rVI<:I ... u ... ""v.L ........... """'..... .Ltv ........ """.1..1..1..1.'-""" .1..1.... V...L 1:-".1.. ....... ...... ~ ~v.L .I..l.I.A/.L.J <:I,.,rl &.400.1..1..\,..1. secondary responsibilities, and may be thought of as two independent but complementary transmission paths superimposed on a single physical circuit. The use of this data link control procedure ensures that packets provided by the packet level and contained in HDLC information frames are accurately exchanged between the DTE and the Network. The functions performed by the Frame Level Interface are: 1. transfer of data in an efficient and timely fashion; 2. synchronizing the link to ensure that the receiver is in step with the transmitter; 3. detecting transmission errors and taking steps to recover from such errors; 4. identifying and reporting procedural errors to higher levels for recovery. The major significance of the Frame Level Interface is that it provides the Packet Level Logical Interface with an error-free, variable delay link behveen the DTE and the Network. The Packet Level Logical Interface is the highest level of the International Packet-Mode Interface and specifies the manner in which control information and user data are structured into packets. The control information including addressing information is contained in the packet header field and allows the network to identify the DTE for which the packet is destined. It also allows a single physical circuit to support a number of virtual circuits to numerous other DTEs concurrently. The Packet Level Logical Interface is further described in the next section . THE PACKET LEVEL LOGICAL INTERFACE Multiplexing at the packet level DTE (customer Bide of DTE/DCE INTERFACE DTE/DCE interface) Nole: DCE (network side 9f DTE/DCE inlerfsce) Network Is tran,pare·nl to proce •• - to - proce s. communication. DC It. Data Circuit - TerminaU,,- Equipment Ce.l. data set) Figure 2-International packet-mode interface architecture The Packet Level Logical Interface accommodates both permanent and switched virtual circuits. A permanent virtual circuit is a permanent association existing between two DTEs which is analogous to a pointto-point private line. Thus, it requires no call set up or call clearing action by the DTE. A switched virtual circuit is a temporary association between two DTEs and is initiated by a DTE sending a call request packet to the network. Call establishment and clearing is described in the next section. From the collection of the Computer History Museum (www.computerhistory.org) 480 National Computer Conference; 1976 In order to allow a DTE to establish concurrent virtual circuits with a number of DTEs over a single physical access circuit, the Packet Level Logical Interface employs packet-interleaved Statistical Multiplexing. This multiplexing technique is used to exploit the fact that a typical virtual circuit to a remote DTE may actually be carrying data for only a small percentage of the t:me. Each packet contains a logical channel number which identifies the packet with a switched or permanent virtual circuit for both directions of transmission. A packet that contains user data for example has a three octet header identifying it as a data packet and specifying its logical channel number as illustrated in Figure 3. BIT ° OCTET 1 2 3 5 6 7 l /--0_ _ 0 _ _0_ _ 1 _L..i _l_0.9_iC_Q' _ _ _ _ _ _ -' 2 Channe I Number 3 Type" Call Request 4 Calling OTE Address Length 5 4 I Called OTE Address length 1, .. ..¥ U o 0.. DTE Address Field Facility Field Length Call establishment and clearing A signalling method is provided to allow a DTE to establish switched virtual circuits to other DTEs using logical channel numbers at each end to locally designate these switched virtual circuits. A DTE initiates a call by sending a call request packet, Figure 4, to the Network. The call request packet includes the logical channel number chosen by the DTE to be used to identify all packets associated with that call. It also includes the network address of the called DTE. A facility field is present only when the DTE wishes to request an optional user facility (i.e., network feature) requiring some indication at call set up. Reverse charging is an example of such a facility. User data may follow the facility field and may contain any number of bits up to a maximum of 16 octets. The calling DTE will receive a response indicating whether or not the ('~lpn nTR ~(,p:rh th,,- ('~n Wh,,-!'l a switched virtual circuit cannot be established, the network will transfer clearing call progress signals Facility Field User Data (0 - 16 octets) Bits of an octet are numbered 0 to 7, where bit 7 is the low order bit and is transmitted first.' Octets of a packet are numbered consecutively starting from 1 and are transmitted in this order. E very packet header has a 4 - bit field which is effectively reserved for future use, a 12 - bit logical channel number ond an 8 - bit field used for packet type information and control. This latter field is intentionally kept very similar to the control field of HDlC. Figure 4-Call request packet fOl'Inat BIT ° OCTET 1 4 3 2 ...-_0_-_0_-0_ _ _---J1 ,_l~i 2 Channel Number peR) I M I 5 pes) Q - Data Qualifier = 7 _______ _ User Data Field M 6 More Data Indicator Figure 3-Data packet format I Type ~ IOata to the DTE indicating the reason why the call was not established; call progress signals are listed in Table I. Either DTE may clear an established call with this information being conveyed to the opposite DTE. Figure 5 is an illustration of call establishment, data transfer and call clearing. Data transfer on a virtual circuit Data packets, illustrated in Figure 2, can only be transferred on a virtual circuit after the virtual circuit has been established and flow control constraints are not violated. The third octet of the data packet header is identical to the control byte of HDLC information frames except that the poll/final bit is replaced by the More Data bit discussed later. From the collection of the Computer History Museum (www.computerhistory.org) International Packet-Mode Interface T ABLE I-Clearing Call Progress Signals CALLII'IG DTE/DCE 481 CALLED DCE/DTE IHTE~FAC INTERP.CE Clearing Call Progress Explanation Signal The called number is fully engaged and cannot accept anothel' call. Number Busy Number Refusing Collect Calls The called DTE does not accept collect calls. Network Congestion Congestion conditions within the network Leinporarily pi'event the requested virtual circuit from being established. Invalid Call Invalid Facility requested. Access Barred The calling DTE is not permitted to obtain the connection to the called number. Possible reason is incompatible closed user group. Local Procedure Error The call is cleared because of a local procedure error. Remote Procedure Error The call is cleared because of a remote procedure error. Not obtainable The called number is not assigned or is no longer assigned. I CALL REQUEST PACKET . T ------- ESTABLISHMENT PHASE _':-~CT DATA "ACKET 4H~- PACKET + PHlE I DATA _ _PACKET .... ___ DATA P.Ar.KET P...... SE T [STABLISIlM[tIT +I DITA i"~C(£'j CALL CN'I'ECT[D ~_PACKET J-T I . INCQMING CALL -1~ ----................. _ .>-- _ -- DATA PACKET ....... DATA PACKET , ----+-. DATA PHASE -----------..... DATA PACKET CLEAR INDICATION Out of Order The called number is out of order. Possible reasons include (1) DTE not functioning; (2) Subscriber link not functioning; (3) Frame level not in operation. P (S) is the packet send sequence number of the packet. (Only data packets are numbered, modulo 8). The maximum number of sequentially numbered data packets that the DTE may be authorized to transmit, without further authorization from the network, may never exceed seven. The actual maximum value, called the window size W, is set for the virtual circuit either at subscription time or at call set up. Each data packet also carries a packet receive sequence number P (R) which authorizes the transmission of W data packets on this virtual circuit starting with a send sequence number equal to the value of P (R). If the DTE or the network wishes to authorize the transmission of one or more data packets across the interface, but there is no data flow on a given virtual circuit in the reverse direction on which to piggyback this information, it can transmit a Receive Ready (RR) packet. Flow control based on the conveyance of P(R) numbers on a virtual circuit basis ensures that a sending DTE does not transmit data at an average rate which is greater than that at which the receiving DTE can accept that data. The data field of a data packet to be transmitted on a virtual circuit may be any number of bits long up to some maximum value. The latter may be established independently at each end of a virtual circuit. Every network will support a maximum value of 128 octets. It may optionally support other values, possible values P~.CYET DISCONNECTION PHASE C LEAR REQUEST PACKET CLEAR -I~ CONFIRMATION PACKET CLE,\R CONFIRMATION PACKET T DISCON~ET PI'AS[ T Figure 5-lllustration of call establishment, data transfer and call clearing being 16, 32, 64, 255, 256, 512 and 1024. The governing principle is that a virtual circuit is used for the transfer of streams of user bits, where packet size may be chosen in such a way as to locally optimize: access line performance, cost, error rates, queuing delays, throughput, etc. In order to facilitate the segmentation and grouping of the user's data stream into data packets, the user may indicate in a full data packet whether there is a logical continuation of his data in the next data packet on a particular virtual circuit. This he does with the More Data bit (M) indicated in Figure 3. Only a full data packet requires a More Data indication since a partially full packet is treated as if it had the M bit off. The use of the M bit ensures that two communicating DTEs can each operate at their locally selected packet sizes. Two independent mechanisms are provided to transfer control information between a pair of DTEs outside the normal flow of data. The first mechanism transfers control data within the normal flow control and sequencing procedures on a virtual circuit. This is called the data qualifier procedure. The format used in this procedure is identical to the normal data trans- From the collection of the Computer History Museum (www.computerhistory.org) 482 National Comruter Conference, 1976 fer packet except that the "Q" bit is set (see Figure 3). The data transmitted is then interpreted by the receiving DTE. An example of the use of the data qualifier is to transfer device control information such as echoing or packet forwarding ruies and transmission control parameters for device-dependent interfaces on the packet network. The second mechanism bypasses the normal data packet transmission sequence providing non-sequenced interrupt packets. Interrupt packets consist of a short header identifying the logical channel number and a one octet data field. Interrupt packets will be transmitted by the network without waiting for all other packets to be delivered and will be delivered to a DTE even when it is not accepting data packets. They contain neither send nor receive sequence numbers. In this way, interrupt conditions, such as would be generated by the depression of a break key on a keyboard terminal, can be signalled between DTEs without being subject to the flow control imposed on data packets. Error recovery The reset procedure is used to reinitialize the flow control procedure on a given virtual circuit to the state it was in when the virtual circuit was established (i.e., all sequence numbers equal to zero and no data in transit). To reach this state, all data and interrupt packets which may be in transit at the time of resetting are discarded. Reset packets are used in the reset procedure. The restart procedure is primarily used by the DTE and provides a mechanism to recover from major failures. The issuance of a restart request packet is a clear re~nf:,t nn R 11 ~wit('hpn eouivalent to ~enrlim virtual circuits and a reset request on all permanent virtual circuits. Thus, the restarting procedure will bring the user/network interface to the state it was in when service was initiated. CONCLUSIONS This paper has presented a description of a new communications protocol for accessing packet switching networks. The International Packet-Mode Interface has been developed by a number of administrations and common carriers in cooperation with standards organizations, users and manufacturers. We strongly believe that it is to the advantage of teleprocessing users, manufacturers of data processing equipment and common carriers that standards and recommendations continue in this area. The benefits to accrue are simplified design and use of equipment, lower cost, higher transmission efficiency, interconnectivity and enhanced performance. The International Packet-Mode Interface has already been implemented in a number of installations to give access to the Telenet, Datapac and experimental RCP networks. It has been specified for use on the French PTT's Transpac network and for the Euronet international network being developed by a large number of European administrations. REFERENCES W. W., Advances in Computer Communications, Artech House Inc., Dedham, Massachusetts, 1974. 2. Cotton, 1. W. and J. W. Benoit, "Prospects for the Standardization of Packet-Switched Networks," Fourth Data Communications Symposium, Quebec City, Canada, October 7-9,1975. 1. Chu, ("'flT,.~ u" ~ VV.J...J....i.. ~"'I.&u) ........1. ,...,.. _ _ 'l.J.LU"""tJ .... .,.TT,.... j ., 1· "..L..I.. \.JUH .... .L.H..H.U,iUll ,.... .... 'IIu. .... .... _ 'T"'\ ",",U"",, .LJtleu~ , 1975. Submitted by the French PTT and U.K. Post Office. From the collection of the Computer History Museum (www.computerhistory.org)