电子工程师手册（英文版四）：ch072

Huber, M.N., Daigle, J.N., Bannister, J., Gerla, M., Robrock II, R.B. “Networks” The Electrical Engineering Handbook Ed. Richard C. Dorf Boca Raton: CRC Press LLC, 2000 72 Networks 72.1 B-ISDN B-ISDN Services and Applications?Asynchronous Transfer Mode?Transmission of B-ISDN Signals?ATM Adaptation Layer?B-ISDN Signaling 72.2 Computer Communication Networks General Networking Concepts?Computer Communication Network Architecture?Local-Area Networks and Internets?Some Additional Recent Developments 72.3 Local-Area Networks The LAN Service Model?Other Features?The Importance of LAN Standards 72.4 The Intelligent Network A History of Intelligence in the Network?The Intelligent Network?Intelligent Network Systems?The CCS7 Network?The Service Control Point?Data Base 800 Service?Alternate Billing Services?Other Services?The Advanced Intelligent Network?Back to the Future 72.1 B-ISDN Manfred N. Huber Since the mid-1980s the idea of the integrated services digital network (ISDN) has become reality. In ISDN voice services with supplementary features and data services with a bit rate of up to 64 kbit/s are integrated in one network. For voice communication and many text and data applications the 64-kbit/s ISDN will be sufficient. Although it is minor as yet, there exists already a growing demand for broadband communication with bit rates from some megabits per second up to approximately 130 Mbit/s [Wiest, 1990] (e.g., high-speed data communication, video communication, high-resolution graphics). In order to provide the same advantages of ISDN to broadband communication users, network operators, and service providers, the development of an intelligent broadband-ISDN (B-ISDN) is necessary. The future B-ISDN will become the universal network integrating different kinds of services with their individual features and requirements. B-ISDN will support switched, semipermanent and permanent, point-to-point, and point- to-multipoint connections and provide on-demand, reserved, and permanent services. B-ISDN connections support packet mode and circuit mode services of mono- and/or multimedia type of a connection-oriented or connectionless nature in a unidirectional or bidirectional configuration [H?ndel and Huber, 1991b]. B-ISDN Services and Applications As already mentioned, there exists some demand for broadband communication which originates from business customers as well as residential customers. In the residential area, on the one hand, people are interested in video distribution services for entertainment purposes, like television and high-definition TV; on the other Manfred N. Huber Siemens J. N. Daigle University of Mississippi Joseph Bannister University of Southern California Information Sciences Institute Mario Gerla University of California, Los Angeles Richard B. Robrock II Bell Communications Research ? 2000 by CRC Press LLC hand, they will use video telephony with acceptable quality. Over the long term, video mail services and video retrieval services will become more important. Voice and text are no longer sufficient for business customers. In the offices and factories of tomorrow, interactive broadband services will be required. Handling complex tasks in the future demands comprehensive support by services for voice, text, data, graphics, video, and documents. In addition to the individual services, the multimedia services and the simultaneous or alternating use of several services with multifunction work- stations will gain importance [Armbrüster, 1990]. Interconnection of local-area networks (LANs) or large computers, computer-aided design, and computer- aided manufacturing will become important data applications. The first video services will be video telephony and video conferencing (studio-to-studio and workstation video conferencing). Initially these services may have diminished quality, but for the long term TV quality can be expected. The bit rates of all services mentioned above are in the range of 2 to 130 Mbit/s (depending on the individual application). Taking into account that in the future more enhanced video coding mechanisms will be available, the required bit rates for video services will become lower without influencing quality significantly. Asynchronous Transfer Mode In today’s public switched networks the synchronous transfer mode (STM) predominates. Applying STM technology, for the duration of a connection a synchronous channel with constant bit rate is allocated to that connection. STM does not fit very well for the integration of services with bit rates from some kilobits per second to 130 Mbit/s. Therefore, in B-ISDN a new transfer mode called asynchronous transfer mode (ATM) is used. In ATM all kinds of information is transported in cells. A cell is a block of fixed length, which consists of a 5-octet cell header and a 48-octet cell payload (see Fig. 72.1). The cell header contains all necessary information for transferring the cell through the network and the cell payload includes the user information. The cell rate of a connection is proportional to the service bit rate. Only if information is available is a cell used by the connection. By having different routing labels, cells of different connections can be transported on the same transmission line (cell multiplexing). If no connection has information ready to transport, idle cells will be inserted. Idle cells do not belong to any connection; they are identified by a standardized cell header. ATM uses only cells; multiplexing and switching of cells is independent of the applications and of the bit rates of the individual connections. Applying ATM technology, the idea of one universal integrated network becomes a reality. However, the ATM technology also causes some problems. Because of the asynchronous multiplexing buffers are necessary, which results in cell delay, cell delay variation, and cell loss. In order to compensate for these effects additional measures have to be provided. FIGURE 72.1 ATM principle. ? 2000 by CRC Press LLC Figure 72.1 also shows the individual subfields of the cell header. The first field, called generic flow control (GFC), is only available at the user-network interface (UNI). Its main purpose is media access control in shared medium configurations (LAN-like configurations) within the customer premises [G?ldner and Huber, 1991]. The proposed GFC procedures are based either on the distributed queue algorithm or the reset timer control mechanism [H?ndel and Huber, 1991a]. At the network-node interface (NNI) these bits are part of the virtual path identifier (VPI). The VPI together with the virtual channel identifier (VCI) form the routing label (identifier of the connec- tion). The VPI itself marks only the virtual path (VP). The VP concept allows the flexible configuration of individual subnetworks (e.g., signaling network or virtual private network), which can be independent of the underlying transmission network. VP networks are under the control of network management. The bandwidth of a VP will be allocated according to its requirements. Within the VP network the individual connections are established and cleared down dynamically (by signaling). The payload type field in the cell header differentiates the information in the cell payload of one connection (e.g., user information, operation and maintenance information for ATM). The value of the cell loss priority bit distinguishes cells that can be discarded under some exceptional network conditions without disturbing the quality significantly from those cells that may not be discarded. The last field of the cell header forms the header error control field. The cell header is protected against errors with a mechanism that allows the correction of a single bit error and the detection of multibit errors. The high transmission speeds for ATM cell transfer require very high-performance switching nodes. There- fore, the switching networks (SNs) have to be implemented in fast hardware. Within the SN the self-routing principle will be applied [Schaffer, 1990]. At the inlet of the SN the cell is extended by an SN-internal header. It is evident that the SN-internal operational speed has to be increased. When passing the individual switching elements, for the processing of the SN-internal header only simple hard-wired logic is necessary. This reduces the control complexity and provides a better failure behavior. When starting several years ago with the imple- mentation of the ATM technology, only the emitter coupled logic (ECL) was available. Nowadays, the comple- mentary metal-oxide semiconductor (CMOS) technology with its low power consumption is used [Fischer et al., 1991]. Transmission of B-ISDN Signals Transmission systems at the UNI provide bit rates of around 150 and 622 Mbit/s. In addition to these rates, at the NNI around 2.5 Gbit/s and up to 10 Gbit/s will be used in the future [Baur, 1991]. In addition to the high- capacity switching and multiplexing technology, high-speed transmission systems are required. Optical fibers are especially suitable for this purpose; however, for the lower bit rates coaxial cables can be used. Optical transmission uses optical fibers as the transmission medium in low-diameter and low-weight cables to provide large transmission capacities over long distances without the need for repeaters. Optical transmission equipment currently tends to mono-mode fiber and laser diodes with wavelengths of around 1310 nm. For both directions in a transmission system either two separate fibers or one common fiber with wavelength division multiplexing can be used. The second solution may be a good alternative for subscriber lines and short trunk lines [Bauch, 1991]. For ATM cell transmission, two possibilities exist, which are shown in Fig. 72.2: synchronous pulse frame or continuous cell stream (cell-based). The basis for the pulse frame concept is the existing synchronous digital hierarchy (SDH). In SDH the cells are transported within the SDH payload; the frame overhead includes operation and maintenance (OAM) of the transmission system. In the cell-based system the OAM for the transmission system is transported within cells. The SDH solution is already defined, whereas for cell-based transmission some problems remain to be solved (e.g., OAM is not yet fully defined). ATM Adaptation Layer The ATM adaptation layer (AAL) is between the ATM layer and higher layers. Its basic function is the enhanced adaptation of the services provided by ATM to the requirements of the layers above. In order to minimize the number of AAL protocols, the service classification shown in Fig. 72.3 was defined. This classification was made with respect to timing relation, bit rate, and connection mode. ? 2000 by CRC Press LLC The AAL protocols are subdivided into two parts. The lower part performs, at the sending side, the segmen- tation of long messages into the cell payload and, at the receiving side, reassembly into long messages. The upper part is service dependent and provides the AAL service to the higher layer. B-ISDN Signaling For signaling in B-ISDN, existing protocols and infrastructure will be reused as much as possible. Figure 72.4 shows the protocol stacks for UNI and NNI. The upper part concerns signaling applications and the lower part signaling transfer. For the introduction of simple switched services in B-ISDN, at UNI and NNI, existing signaling application protocols will be reused. The 64-kbit/s ISDN-specific information elements will be removed and new B-ISDN- specific information elements will be added. Right from the beginning these protocols will provide means that allow smooth migration toward future applications, which will include highly sophisticated features like mul- timedia services [Huber et al., 1992]. This approach guarantees compatibility for future protocol versions. At the NNI the existing signaling system no. 7 (SS7) can be reused (see right part of the NNI protocol stack in Fig. 72.4). SS7 is a powerful and widespread network that will continue to be applied for rather a long period until ATM penetration has been reached. For the middle term, however, a fully ATM-based network will be available which also carries signaling messages (see left part of the NNI protocol stack in Fig. 72.4). ATM-based signaling at the NNI needs a suitable AAL which provides the services of the existing message transfer part level 2. At the UNI, right from the beginning, all kinds of traffic (including signaling) is carried within cells. An AAL for signaling at the UNI is also required. This AAL has to provide the services of the existing layer 2 UNI FIGURE 72.2 Transmission principles for B-ISDN. FIGURE 72.3 AAL service classification. ? 2000 by CRC Press LLC protocol. The AAL for signaling at UNI and NNI will be common as much as possible. In contrast to the NNI, at the UNI meta-signaling is necessary. Meta-signaling establishes, checks, and removes the signaling channels between customer equipment and the central office in a dynamic way. The signaling channels at the NNI are semipermanent and, therefore, meta-signaling is not required. Defining Terms Asynchronous transfer mode: A transfer mode in which the information is organized into cells; it is asyn- chronous in the sense that the recurrence of cells containing information from an individual user is not necessarily periodic. ATM adaptation layer: A layer which provides the adaptation of higher layers to ATM. Broadband: A service or system requiring transmission channels capable of supporting bit rates greater than 2 Mbit/s. Cell: A block of fixed length which is subdivided into a cell header and an information field. The cell header contains a label which allows the clear allocation of a cell to a connection. Integrated services digital network: A network which provides end-to-end digital connectivity to support a wide range of services, including voice and nonvoice services, to which users have access by a limited set of standard multipurpose user-network interfaces. Signaling: Procedures which are used to control (set up and clear down) calls and connections within a telecommunication network. Synchronous digital hierarchy: A standard for optical transmission which provides transmission facilities with flexible add/drop capabilities to allow simple multiplexing and demultiplexing of signals. Related Topic 72.2 Computer Communications Networks FIGURE 72.4 Protocol stacks for B-ISDN signaling. ? 2000 by CRC Press LLC References H. Armbrüster, “Blueprint for future telecommunications,” Telcom Report International, vol. 13, no. 1, pp. 5–8, 1990. H. Bauch, “Transmission systems for B-ISDN,” IEEE LTS, Magazine of Lightwave Telecommunication, vol. 2, no. 3, pp. 31–36, 1991. H. Baur, “Technological perspective of telecommunications for the nineties,” Integration, Interoperation and Interconnection: This Way to Global Services, Proceedings of the Technical Symposium, Geneva, part 2, vol. 1 paper 1.1, 1991. W. Fischer, O. Fundneider, E.-H. Goeldner, and K.A. Lutz, “A scalable ATM switching system architecture,” IEEE Journal on Selected Areas in Communication, vol. 9, no. 8, pp. 1299–1307, 1991. E.-H. G?ldner and M.N. Huber, “Multiple access for B-ISDN,” IEEE LTS, Magazine of Lightwave Telecommu- nication, vol. 2, no. 3, pp. 37–43, 1991. R. H?ndel and M.N. Huber, “Customer network configurations and generic flow control,” International Journal of Digital and Analog Communication Systems, vol. 4, no. 2, pp. 117–122, 1991a. R. H?ndel and M.N. Huber, Integrated Broadband Networks — An Introduction to ATM-Based Networks, Reading, Mass.: Addison-Wesley, 1991b. M.N. Huber, V. Frantzen, and G. Maegerl, “Proposed evolutionary paths for B-ISDN signalling,” Proceedings of the XIV International Switching Symposium, Yokohama, vol. 1, pp. 334–338, 1992. B. Schaffer, “ATM switching in the developing telecommunication network,” Proceedings of the XIII International Switching Symposium, vol. 1, pp. 105–110, 1990. G. Wiest, “More intelligence and flexibility for communication network—Challenges for tomorrow’s switching systems,” Proceedings of the XIII International Switching Symposium, vol. 5, pp. 201–204, 1990. Further Information CCITT Recommendations and CCITT Draft Recommendations concerning B-ISDN (parts of F, G, I and Q series), which are published by the International Telecommunication Union. Journals of the IEEE Communication Society (Communications Magazine, Journal on Selected Areas in Communications, LTS: Magazine of Lightwave Telecommunication, Networks, Transactions on Communications), which are published by the Institute of Electrical and Electronics Engineers, Inc. International Journal of Digital and Analog Communication System, which is published by John Wiley & Sons, Ltd. Proceedings of international conferences such as GLOBECOM, INFOCOM, International Conference on Communications, International Conference on Computer Communication, International Switching Sympo- sium, International Symposium on Subscriber Loops and Services, and International Teletraffic Congress. A detailed description of ISDN is given in ISDN—The Integrated Services Digital Network— Concepts, Methods, Systems, by P. Bocker, published by Springer-Verlag. 72.2 Computer Communication Networks J. N. Daigle A computer communication network is a collection of applications hosted on different machines and inter- connected by an infrastructure that provides communications among the communicating entities. While the applications are generally understood to be computer programs, the generic model includes the human being as an application. In fact, one or all of the “applications’’ that are communicating may be human beings. This section summarizes the major characteristics of computer communication networks. The objective is to provide a concise introduction that will allow the reader to gain an understanding of the key distinguishing characteristics of the major classes of networks that exist today and some of the issues involved in the intro- duction of emerging technologies. There are a significant number of well-recognized books in this area. Among these are the excellent texts by Schwartz [1987], Tanenbaum [1988], and Spragins [1991], which have enjoyed wide acceptance by both students and practicing engineers and cover most of the general aspects of computer communication networks. Stallings ? 2000 by CRC Press LLC [1990a, 1990b, 1990c] covers a broad array of standards in this area. Other books that have been found to be especially useful by practitioners are those by Rose [1990] and Black [1992]. The latest developments are, of course, covered in the current literature, conference proceedings, and the notes of standards meetings. A pedagogically oriented magazine that specializes in computer communications networks is IEEE Network, but IEEE Communications and IEEE Computer often also contain interesting articles in this area. ACM Communications Review, in addition to presenting pedagogically oriented articles, often presents very useful summaries of the latest standards activities. Major conferences that specialize in computer communications include the IEEE INFOCOM and ACM SIGCOMM series, which are held annually. We will begin our discussion with a brief statement of how computer networking came about and a capsule description of the networks that resulted from the early efforts. Networks of this generic class, called wide-area networks (WANs), are broadly deployed today, and there are still a large number of unanswered questions with respect to their design. The issues involved in the design of those networks are basic to the design of most networks, whether wide area or otherwise. In the process of introducing these early systems, we will describe and contrast three basic types of communication switching: circuit, message, and packet. We will next turn to a discussion of computer communication architecture, which describes the structure of communication-oriented processing software within a communication processing system. Our discussion is limited to the International Standards Organization/Open Systems Interconnection (ISO/OSI) reference model (ISORM) because it provides a framework for discussion of some of the modern developments in communications in general and communication networking in particular. This discussion is necessarily sim- plified in the extreme, thorough coverage requiring on the order of several hundred pages, but we hope our brief description will enable the reader to appreciate some of the issues. Having introduced the basic architectural structure of communication networks, we will next turn to a discussion of an important variation on this architectural scheme: the local-area network (LAN). Discussion of this topic is important because it helps to illustrate what the reference model is and what it is not. In particular, the architecture of LANs illustrates how the ISORM can be adapted for specialized purposes. Specifically, early network architectures anticipate networks in which individual node pairs are interconnected via a single link, and connections through the network are formed by concatenating node-to-node connections. LAN architectures, on the other hand, anticipate all nodes being interconnected in some fashion over the same communication link (or medium). This, then, introduces the concept of adaption layers in a natural way. It also illustrates that if the services provided by an architectural layer are carefully defined, then the services can be used to implement virtually any service desired by the user, possibly at the price of some inefficiency. After discussing LANs, we will conclude our article with a discussion of two of the variants in packet switching transmission technology: frame relay and a recent development in basic transmission technology called the asyn- chronous transfer mode, which is a part of the larger broadband integrated services digital network effort. These technologies are likely to be important building blocks for the computer communication networks of the future. General Networking Concepts Data communication networks have existed since about 1950. The early networks existed primarily for the purpose of connecting users of a large computer to the computer itself, with additional capability to provide communications between computers of the same variety and having the same operating software. The lessons learned during the first twenty or so years of operation of these types of networks have been valuable in preparing the way for modern networks. For the purposes of our current discussion, however, we will think of commu- nication networks as being networks whose purpose is to interconnect a set of applications that are implemented on hosts manufactured by possibly different vendors and managed by a variety of operating systems. Networking capability is provided by software systems that implement standardized interfaces specifically designed for the exchange of information among heterogeneous computers. During the late 1960s, many forward-looking thinkers began to recognize that significant computing resources (that is, supercomputers) would be expensive and unlikely to be affordable by many of the researchers needing this kind of computer power. In addition, they realized that significant computing resources would not be needed all of the time by those having local access. If the computing resource could be shared by a number of research sites, then the cost of the resource could be shared by its users. ? 2000 by CRC Press LLC Many researchers at this time had computing resources available under the scenario described in the first paragraph above. The idea of interconnecting the computers to extend the reach of these researchers to other computers developed. In addition, the interconnection of the computers would provide for communication among the researchers themselves. In order to investigate the feasibility of providing the interconnectivity anticipated for the future using a new technology called packet switching, the Advanced Research Projects Agency (ARPA) of the Department of the Army sponsored a networking effort, which resulted in the computer communication network called the ARPANET. The end results of the ARPA networking effort, its derivatives, and the early initiatives of many companies such as AT&T, DATAPOINT, DEC, IBM, and NCR have been far-reaching in the extreme. Any finitely delimited discussion of the accomplishments of those efforts would appear to underestimate their impact on our lives. We will concentrate on the most visible product of these efforts, which is a collection of programs that allows applications running in different computers to intercommunicate. Before turning to our discussion of the software, however, we will provide a brief description of a generic computer communication network. Figure 72.5 shows a diagram of a generic computer communication network. The most visible components of the network are the terminals, the access lines, the trunks, and the switching nodes. Work is accomplished when the users of the network, the terminals, exchange messages over the network. The terminals represent the set of communication terminating equipment communicating over the network. Equipment in this class includes, but is not limited to, user terminals, general-purpose computers, and database systems. This equipment, either through software or through human interaction, provides the functions required for information exchange between pairs of application programs or between application programs and people. The functions include, but are not limited to, call set-up, session management, and message transmission control. Examples of applications include electronic mail transfer, terminal-to-computer connec- tion for time sharing or other purposes, and terminal-to-database connections. Access lines provide for data transmission between the terminals and the network switching nodes. These connections may be set up on a permanent basis or they may be switched connections, and there are numerous transmission schemes and protocols available to manage these connections. The essence of these connections, however, from our point of view is a channel that provides data transmission at some number of bits per second (bps), called the channel capacity, C. The access line capacities may range from a few hundred bits per second to in excess of millions of bits per second, and they are usually not the same for all terminating equipments of a given network. The actual information-carrying capacity of the link depends upon the protocols employed to effect the transfer; the interested reader is referred to Bertsekas and Gallagher [1987], especially Chapter 2, for a general discussion of the issues involved in transmission of data over communication links. Trunks, or internodal trunks, are the transmission facilities that provide for transmission of data between pairs of communication switches. These are analogous to access lines, and, from our point of view, they simply provide a communication path at some capacity, specified in bits per second. FIGURE 72.5 Generic computer communication network. ? 2000 by CRC Press LLC There are three basic switching paradigms: circuit, message, and packet switching. Circuit switching and packet switching are transmission technologies while message switching is a service technology. In circuit switching, a call connection between two terminating equipments corresponds to the allocation of a prescribed set of physical facilities that provide a transmission path of a certain bandwidth or transmission capacity. These facilities are dedicated to the users for the duration of the call. The primary performance issues, other than those related to quality of transmission, are related to whether or not a transmission path is available at call set-up time and how calls are handled if facilities are not available. Message switching is similar in concept to the postal system. When a user wants to send a message to one or more recipients, the user forms the message and addresses it. The message switching system reads the address and forwards the complete message to the next switch in the path. The message moves asynchronously through the network on a message switch-to-message switch basis until it reaches its destination. Message switching systems offer services such as mail boxes, multiple destination delivery, automatic verification of message delivery, and bulletin board. Communication links between the message switches may be established using circuit or packet switching networks as is the case with most other networking applications. Examples of message switching protocols that have been used to build message switching systems are Simple Mail Transfer Protocol (SMTP) and the International Telegraph and Telephone Consultative Committee (CCITT) X.400 series. The former is much more widely deployed, while the latter has significantly broader capabilities, but its deployment is plagued by having two incompatible versions (1984 and 1988) and other problems. Many commercial vendors offer message switching services based on either one of the above protocols or a proprietary protocol. In the circuit switching case, there is a one-to-one correspondence between the number of trunks between nodes and the number of simultaneous calls that can be carried. That is, a trunk is a facility between two switches that can service exactly one call, and it does not matter how this transmission facility is derived. Major design issues include the specification of the number of trunks between node pairs and the routing strategy used to determine the path through a network in order to achieve a given call blocking probability. When blocked calls are queued, the number of calls that may be queued is also a design question. A packet-switched communication system exchanges messages among users by transmitting sequences of packets which comprise the messages. That is, the sending terminal equipment partitions a message into a sequence of packets, the packets are transmitted across the network, and the receiving terminal equipment reassembles the packets into messages. The transmission facility interconnecting a given node pair is viewed as a single trunk, and the transmission capacity of this trunk is shared among all users whose packets traverse both nodes. While the trunk capacity is specified in bits per second, the packet handling capacity of a node pair depends both upon the trunk capacity and the nodal processing power. In many packet-switched networks, the path traversed by a packet through the network is established during a call set-up procedure, and the network is referred to as a virtual circuit packet switching network. Other networks provide datagram service, a service that allows users to transmit individually addressed packets without the need for call set-up. Datagram networks have the advantage of not having to establish connections before communication takes place, but they have the disadvantage that every packet must contain complete addressing information. Virtual circuit networks have the advantage that addressing information is not required in each packet, but have the disadvantage that a call set-up must take place before communication can occur. Datagram is an example of connectionless service while virtual circuit is an example of connection-oriented service. Prior to the late 1970s, signaling for circuit establishment was in-band. That is, in order to set up a call through the network, the call set-up information was sent sequentially from switch to switch using the actual circuit that would eventually become the circuit used to connect the end users. In an extreme case, this amounted to trying to find a path through a maze, sometimes having to retrace one’s steps before finally emerging at the destination or just simply giving up when no path could be found. This had two negative characteristics: first, the rate of signaling information transfer was limited to the circuit speed, and second, the circuits that could have been used for accomplishing the end objective were being consumed simply to find a path between the end-points. This resulted in tremendous bottlenecks on major holidays, which were solved by virtually disal- lowing alternate routes through the toll switching network. An alternate out-of-band signaling system, usually called common-channel interoffice signaling (CCIS), was developed primarily to solve this problem. Signaling now takes place over a signaling network that is ? 2000 by CRC Press LLC partitioned from the network that carries the user traffic. This principle is incorporated into the concept of integrated services digital networks (ISDNs), which is described thoroughly in Helgert [1991]. The basic idea of ISDN is to offer to the user some number of 64-kbps access lines plus a 16-kbps access line through which the user can describe to an ISDN how the user wishes to use each of the 64-kbps circuits at any given time. The channels formed by concatenating the access lines with the network interswitch trunks having the requested characteristics are established using an out-of-band signaling system, the most modern of which is signaling system #7 (SS#7). In either virtual circuit or datagram networks, packets from a large number of users may simultaneously need transmission services between nodes. Packets arrive at a given node at random times. The switching node determines the next node in the transmission path, and then places the packet in a queue for transmission over a trunk facility to the next node. Packet arrival processes tend to be bursty, that is, the number of packet arrivals over fixed-length intervals of time has a large variance. Because of the burstiness of the arrival process, packets may experience significant delays at the trunks. Queues may also build due to the difference in transmission capacities of the various trunks and access lines. Combining of packets that arrive at random times from different users onto the same line, in this case a trunk, is called statistical multiplexing. In addition to the delays experienced at the input to trunks, packets may also experience queueing delays within the switching nodes. In particular, the functions required for packet switching are effected by executing various software processes within the nodes, and packets must queue while awaiting execution of the various processes on their behalf. Both transmission capacities and nodal processing capabilities are available over a wide range of values. If the trunk capacities are relatively low compared to nodal processing capability, then delays at switching nodes may be relatively small. If line capacities are large compared to nodal processing capabilities, however, delays due to nodal processing may be significant. In the general case, all possible sources of delay should be examined to determine where bottlenecks, and consequently delay, occur. It is often the case that a particular point in the communication network, either a processing node or a trunk, is the primary source of delay. In this case, this point is usually singled out for analysis, and a simple model is invoked to analyze the performance at that point. The results of this analysis, combined with results of other analyses, result in a profile of overall system performance. In this case, the key aspect of the analysis is to choose an appropriate model for the isolated analysis. In this way, a simplified analysis leading to useful results can be performed, and this can lead to an improved network design. Protocol design and performance issues are frequent topics of discussion at both general conferences in communications and those specialized to networking. The reader is encouraged to consult the proceedings of the conferences mentioned earlier for a better appreciation of the range of issues and the diversity of the proposed solutions to the issues. Computer Communication Network Architecture In this section, we will begin with a brief, high-level definition of the ISORM. The reference model has seven layers, none of which can be bypassed conceptually. In general, a layer is defined by the types of services it provides to its users and the quality of those services. For each layer in the ISO/OSI architecture, the user of a layer is the next layer up in the hierarchy, except for the highest layer for which the user is an application. Clearly, when a layered architecture is implemented under this philosophy, then the quality of service obtained by the end user, the application, is a function of the quality of service provided by all of the layers. In order to clarify the communications strategy of the ISO/OSI architecture, we will provide a discussion of the layer 2 services in some detail. There is significant debate over whether the efforts of the ISO/OSI community are leading to the best standards (or even standards that have any merit whatever!). Limiting our discussion to the ISORM is, by no means, an endorsement of the actual protocols that have been developed in the ISO arena; there are actually more widely deployed and successful standards in other arenas. On the other hand, the ISORM is very useful for discussing network architecture principles, and these principles apply across the board. Thus, we choose to base our discussion on the ISORM. ? 2000 by CRC Press LLC Figure 72.6, adopted from Spragins [1991], shows the basic structure of the OSI architecture and how this architecture is envisaged to provide for exchange of information between applications. As shown in the figure, there are seven layers: application, presentation, session, transport, network, data link, and physical. Brief definitions of the layers follow, but the reader should bear in mind that substantial further study will be required to develop an understanding of the practical implications of the definitions: ? Physical layer: Provides electrical, functional, and procedural characteristics to activate, maintain, and deactivate physical data links that transparently pass the bit stream for communication between data link entities. ? Data link layer: Provides functional and procedural means to transfer data between network entities; provides for activation, maintenance, and deactivation of data link connections, character and frame synchronization, grouping of bits into characters and frames, error control, media access control, and flow control. ? Network layer: Provides switching and routing functions to establish, maintain, and terminate network layer connections, and transfer data between transport layers. ? Transport layer: Provides host-to-host, cost-effective, transparent transfer of data, end-to-end flow control, and end-to-end quality of service as required by applications. ? Session layer: Provides mechanisms for organizing and structuring dialogues between application pro- cesses. ? Presentation layer: Provides for independent data representation and syntax selection by each commu- nicating application and conversion between selected contexts and the internal architecture standard. ? Application layer: Provides applications with access to the ISO/OSI communication stack and certain distributed information services. As we have mentioned previously, a layer is defined by the types of services it provides to its users. In the case of a request or a response, these services are provided via invocation of service primitives of the layer in question by the layer that wants the service performed. In the case of an indication or a confirm, these services are provided via invocation of service primitives of the layer in question by the same layer that wants the service performed. This process is not unlike a user of a programming system calling a subroutine from a scientific subroutine package in order to obtain a service, say, matrix inversion or memory allocation. For example, a request is analogous to a CALL statement in a FORTRAN program, and a response is analogous to the RETURN statement in the subroutine that has been CALLed. The requests for services are generated asynchronously by all of the FIGURE 72.6 Layered architecture for ISO/OSI reference model. ? 2000 by CRC Press LLC users of all of the services and these join (typically prioritized) queues along with other requests and responses while awaiting servicing by the processor or other resource such as a transmission line. The service primitives fall into four basic types: request, indication, response, and confirm. These types are defined as follows: ? Request: A primitive sent by layer (N + 1) to layer N to request a service. ? Indication: A primitive sent by layer N to layer (N + 1) to indicate that a service has been requested of layer N by a different layer (N + 1) entity. ? Response: A primitive sent by layer (N + 1) to layer N in response to an indication primitive. ? Confirm: A primitive sent by layer N to layer (N + 1) to indicate that a response to an earlier request primitive has been received. In order to be more specific about how communication takes place, we will now turn to a brief discussion of layer 2, the data link layer. The primitives provided by the ISO data link (DL) layer are as follows [Stallings, 1990a]: DL_CONNECT.request DL_RESET.request DL_CONNECT.indication DL_RESET.indication DL_CONNECT.response DL_RESET.response DL_CONNECT.confirm DL_RESET.confirm DL_DATA.request DL_DISCONNECT.request DL_DATA.indication DL_DISCONNECT.indication DL_DATA.response DL_UNITDATA.request DL_DATA.confirm DL_UNITDATA.indication Each primitive has a set of formal parameters, which are analogous to the formal parameters of a procedure in a programming language. For example, the parameters for the DL_CONNECT.request primitive are the Called Address, the Calling Address, and the Quality of Service Parameter Set. The four primitives are used in the establishment of data link connections. The called address and the calling address are analogous to the telephone numbers of two parties of a telephone call, while the quality of service parameter set allows for the negotiation of various agreements such as throughput measured in bits per second. All four DL_CONNECT primitives are used to establish a data link. An analogy to an ordinary phone call can better illustrate the basic idea of the primitives. The DL_CONNECT.request is equivalent to picking up the phone and dialing. The phone ringing at the called party’s end is represented by the DL_CONNECT.indi- cation. DL_CONNECT.response is equivalent to the called party lifting the receiver and answering, and DL_CONNECT.confirm is equivalent to the calling party hearing the response of the called party. In general, communication takes place between peer layer protocols by the exchange of protocol data units (PDUs), which contain all of the information required for the receiving protocol entity to provide the required service. In order to exchange PDUs, entities at a given layer use the services of the next lower layer. The data link primitives listed above include both connection-mode primitives and connectionless-mode primitives. For connection-mode communications, a connection must be established between two peer entities before they can exchange PDUs. For example, suppose a network layer entity in Host A wishes to be connected to a network layer entity in Host B, as shown in Fig. 72.6. Then the connection would be accomplished by the concatenation of two data link connections: one between A and C, and one between C and B. In order to establish the connection, the network layer entity in Host A would issue a DL_CONNECT.request to its associated data link entity, providing the required parameters. This data link entity would then transmit this request to a data link entity in C, which would issue a DL_CONNECT.indication to a network entity in C. The network entity in C would then analyze the parameters of the DL_CONNECT.indication and realize that the target destination is B. This network layer entity would then reissue the DL_CONNECT.request to its data link entity, which would transmit the request to a data link entity in B. The data link entity in B would send a DL_CONNECT.indication to a network layer entity in B, and this entity would issue a DL_CONNECT.response back to the data link entity in B. This DL_CONNECT.response would be relayed back to the data link entity in A following the same sequence of events as in the forward path. Eventually, this DL_CONNECT.response would be converted to a ? 2000 by CRC Press LLC DL_CONNECT.confirm by the data link entity in A and passed to the network entity in A, thus completing the connection. Once the connection is established, data exchange between the two network layer entities can take place; that is, the entities can exchange PDUs. For example, if a network layer entity in Host A wishes to send a PDU to a network layer entity in Host B, the network layer entity in Host A would issue a DL_DATA.request to the appropriate data link layer entity in Host A. This entity would package the PDU together with appropriate control information into a data link service data unit (DLSDU) and send it to its peer at C. The peer at C would deliver it to the network entity at C, which would forward it to the data link entity in C providing the connection to Host B. This entity would then send the DLSDU to its peer in Host B, and this data link entity would pass the PDU to Host B network entity via a DL_DATA.indication. Network layer PDUs are called packets and data link layer PDUs are called frames. The data link layer does not know that the information it is transmitting is a packet; to the data link layer entity, the packet is simply user information. From the perspective of a data link entity, it is not necessary to have a network layer. The network layer exists to add value for the user of the network layer to the services provided by the data link layer. In the example above, value was added by the network layer by providing a relaying capability since Hosts A and C were not directly connected. Similarly, the data link layer functions on a hop-by-hop basis, each hop being completely unaware that there are any other hops involved in the communication. We will see later that the data link need not be limited to a single physical connection. The philosophy of the ISO/OSI architecture is that in addition to the software being layered, implementations are not allowed to bypass entire layers; that is, every layer must appear in the implementation. This approach was developed after the approach defined for the ARPANET project, which is hierarchical, was fully developed. In the hierarchical approach, the layer interfaces are carefully designed, but any number of layers of software can be bypassed by any application (or other higher-layer protocol) that provides the appropriate functionality. These two approaches have been hotly debated for a number of years, but as the years pass, the approaches are actually beginning to look more and more alike for a variety of reasons that will not be discussed here. The ISO/OSI layered architecture described above would appear to be very rigid, not allowing for any variations in underlying topology or variations in link reliability. However, as we shall see, this is not necessarily the case. As an example, ISO 8348, which developed as a result of the X.25 project, provides only connection- oriented service, and it was originally intended as the only network layer standard for ISO/OSI. However, ISO 8473, or ISO-IP, which is virtually identical to the Department of Defense (DoD) internet protocol (DoD-IP) developed in the ARPANET project, has since been added to the protocol suite to provide connectionless service as well as internet service. An interesting aside is that because of the addressing limitations of DoD-IP, the Internet Administrative Board (IAB) has recently recommended replacement of the DoD-IP protocol by the ISO-IP protocol, thus bringing the process full circle. The ISO/OSI protocol suite is in a constant state of revision as new experience reveals the need for additional capabilities and flexibility. Some of this additional flexibility and functionality is being provided through the use of so-called adaption sublayers, which enhance the capabilities of a given layer so that it can use the services of a lower layer with which it was not specifically designed for compatibility. Interestingly, the use of adaption sublayers is only a short step away from using adaption layers that would allow applications to directly interface with any ISO layer. This would result in a hierarchical rather than layered architecture; to wit: ISORM becomes DoDRM. Indeed, fundamental changes in the national (and worldwide) communications infrastructure appear to be leading naturally in the hierarchical direction. Of course, the indiscriminate use of such adaptions would lead back to the proliferation of incompatible protocols and interfaces, the frustration that led to the current twenty-year standardization crusade! It is refreshing to note that a return to our former state does not appear to be around the corner; most standardization work is actually headed in the direction of allowing open systems to intercommunicate. Local-Area Networks and Internets We will now turn to a discussion of LANs, which have inherent properties that make the use of sublayers particularly attractive. In this section, we will discuss the organization of communications software for LANs. In addition, we will introduce the idea of internets, which were brought about to a large extent by the advent ? 2000 by CRC Press LLC of LANs. We will discuss the types of networks only briefly and refer the reader to the many excellent texts on the subject. Layers 4 and above for local-area communications networks are identical to those of wide-area networks. However, because the hosts communicating over a LAN share a single physical transmission facility, the routing functions provided by the network layer, layer 3, are not necessary. Thus, the functionality of a layer 3 in a LAN can be substantially simplified without loss of utility. On the other hand, a data link layer entity must now manage many simultaneous data link layer connections because all connections entering and leaving a host on a single LAN do so over a single physical link. Thus, in the case of connection-oriented communications, the software must manage several virtual connections over a single physical link. There were several basic types of transmission schemes in use in early LANs. Three of these received serious consideration for standardization: the token ring, token bus, and carrier-sense multiple access (CSMA). In a token ring network, the stations are configured on a physical ring around the medium. A token rotates around this physical ring, visiting each host (or station) in turn. A station wishing to transmit data must wait until the token is available to that station. In a token bus LAN, the situation is the same, except that the stations share a common bus and the ring is logical rather than physical. In a CSMA network, the stations are bus connected, and a station may transmit whenever other stations are not currently transmitting. That is, a station wishing to transmit senses the channel, and if there is no activity, the station may transmit. Of course, the actual access protocol is significantly more complicated than this. In the early 1980s, there was significant debate over which LAN connection arrangement was superior, a single choice being viewed as necessary. This debate centered on such issues as cost, network throughput, network delay, and growth potential. Performance evaluation based on queueing theory played a major role in putting these issues in perspective. For thorough descriptions of LAN protocols and queueing models used to evaluate their performance, the interested reader is referred to Hammond and O’Reilly [1986]. All three of the access methods mentioned above became IEEE standards (IEEE 802) and eventually became ISO standards (ISO 8802 series) because all merited standardization. On the other hand, all existed for the express purpose of exchanging information among peers, and it was recognized at the outset that the upper end of the data link layer could be shared by all three access techniques. On the other hand, the lower-level functions of the layer deal with interfacing to the physical media. Here, drastic differences in the way the protocol had to interface with the media were recognized. Thus, a different media-access control sublayer (MAC) was needed for each of the access techniques. The decision to use a common logical link control (LLC) sublayer for all of the LAN protocols apparently ushered in the idea of adaption sublayers. The reason for splitting the layer is simple: a user of the data link control (DLC) layer need not know what kind of medium provides the communications; all that is necessary is that the user understand the interface to the DLC layer. On the other hand, the media of the three types of access protocols provide transmission service in different ways, so software is needed to bridge the gap between what the user of the service needs, which is provided by the LLC, and how the LLC uses the media to provide the required service. Thus, the MAC sublayer was born. This idea has proven to be valuable as new types of technologies have become available. For example, the new fiber-distributed digital interface (FDDI) uses the LLC of all other LAN protocols, but its MAC is completely different from the token ring MAC even though FDDI is a token ring protocol. Reasons for needing a new MAC for LLC are provided in Stallings [1990b]. One of the more interesting consequences of the advent of local-area networking is that many traditional computer communication networks became internets overnight. LAN technology was used to connect stations to a host computer, and these host computers were already on a WAN. It was then a simple matter to provide a relaying, or bridging, service at the host in order to provide wide-area interconnection of stations to LANs to each other. In short, the previously established WANs became networks for interconnection of LANs; that is, they were interconnecting networks rather than stations. Internet performance suddenly became a primary concern in the design of networks. More recently, FDDI is being thought of as a mechanism to provide LAN interconnection on a site basis, and a new type of network, the metropolitan-area network (MAN) has been under study for the intercon- nection of LANs within a metropolitan area. The primary media configuration for MANs is a dual bus configuration and it is implemented via the distributed queue, dual bus (DQDB) protocol, also known as IEEE 802.6. The net effect of this protocol is to use the dual bus configuration to provide service approaching the ? 2000 by CRC Press LLC first-come–first-served service discipline to the traffic entering the FDDI network, which is remarkable con- sidering that the LANs being interconnected are geographically dispersed. Interestingly, DQDB concepts have recently been adapted to provide wide-area communications. Specifically, structures have been defined for transmitting DQDB frames over standard DS-1 (1.544 megabits per second [Mb/s]) and DS-3 (6.312 Mb/s) facilities, and these have been used as the basis for a new service offering called switched multimegabit data services (SMDS). As of this writing, advances in LANs design and new forms of LANs are emerging. One example is wireless LANs, which are LANs in which radio or photonic links serve as cable replacements. Wireless LAN technology is viewed by many as crucial to the evolution of personal communication networks. Another example is the asynchronous transfer mode-based LAN, which is mentioned in the next section following a general discussion of asynchronous transfer mode. Some Additional Recent Developments In this subsection, we will describe two recent developments of significant interest in communication network- ing: fast packet networks and transmission using the asynchronous transfer mode (ATM), which is a part of the larger broadband integrated services digital network (B-ISDN) effort. As we have mentioned previously, there is really no requirement that the physical media between two adjacent data link layers be composed of a single link. In fact, if a path through the network is initially established between two data link entities, there is no reason that DLC protocols need to be executed at intermediate nodes. Figure 72.7, adapted from Bhargava and Hluchyj [1990] shows how the end-to-end connection might be implemented. A network implemented in the fashion indicated in Fig. 72.7 is called a fast packet network (FPN). From Fig. 72.7, it is seen that the data link layer is partitioned into three sublayers: the data link control sublayer (which parallels the LLC layer of LANs), the fast packet adaption (FPA) sublayer, and the fast packet relay (FPR) sublayer. The function of the fast packet adaption sublayer is to segment the layer-2 PDU, the frame, into smaller units, called fast packets, for transmission over the FPN. These fast packets contain infor- mation that identifies the source and destination node names and the frame to which they belong so that they can be routed through the network and reassembled at the destination. The fast packets are statistically multiplexed onto a common physical link by the FPR sublayer for transmis- sion. At intermediate nodes, minor error checking, fast packet framing, fast packet switching, and queueing takes place. If errors are found, then the fast packet is dropped. When the fast packets reach their destination, they are reassembled into a frame by the FPA sublayer and passed on to the DLC sublayer where normal DLC functions are performed. The motivation for FPNs is that since link transmission is becoming more reliable, extensive error checking and flow control are not needed across individual links; an end-to-end check should be sufficient. Meanwhile, the savings in processing due to not processing at the network layer can be applied to frame processing, which allows interconnection of the switches at higher line speeds. Since bits-per-second costs decrease with increased line speed, service providers can offer savings to their customers through FPNs. Significant issues are fast packet loss probability and retransmission delay. Such factors will determine the retransmission strategy deployed in the network. Of course, the goal is to improve FIGURE 72.7Fast packet switched layered architecture. ? 2000 by CRC Press LLC network efficiency, so a significant issue is whether FPNs are better than ordinary packet networks and, if so, by how much. Another recent innovation is the ATM, usually associated with B-ISDN. The idea of ATM is to partition a user’s data into many small segments, called cells, for transmission over the network. Independent of the data’s origin, the cell size is 53 octets, of which 5 octets are for use by the network itself for routing and error control. Users of the ATM are responsible for segmentation and reassembly of their data. Any control information required for this purpose must be included in the 48 octets of user information in each cell. In the usual case, these cells would be transmitted over networks that would provide users with 135 Mb/s and above data transmission capacity (with user overhead included in the capacity). The segmentation of units of data into cells introduces tremendous flexibility for handling different types of information, such as voice, data, image, and video, over a single transmission facility. As a result, LANs, WANs, and MANs based on the ATM paradigm are being designed, and indeed deployed. A significant portion of the deployment activity is a national testbed program, which involves industrial/academia cooperation, under joint sponsorship of the National Science Foundation (NSF) and the Defense Advanced Research Projects Agency (DARPA). There is also significant private investment in developing this technology; for example, experimental ATM-based LANs are already in place at the Digital Equipment Corporation (DEC) research facility in Palo Alto, California. There is some possibility that LANs of this type, rather than of the FDDI type, will be the dominant means of providing high-speed LAN and LAN-interconnect services. There are numerous possibilities for connection of hosts to ATM networks, but they all share a common architecture, which consists of three sublayers: the ATM adaption layer (AAL), the ATM layer, and the physical media-dependent (PMD) layer. Connection of hosts to ATM at a given layer is achieved through developing an AAL for the layer in question. For example, one might decide to adapt to ATM at the network layer. In that case, the transport layer would operate as usual, and the AAL would be designed to process data structures from the transport layer to produce data structures for use by the ATM layer and vice versa. Of course, all hosts communicating with each other in this way would use the same AAL. Figure 72.8 shows an example of how an ISO/OSI host might connect to an ATM network. Below the data link layer is the ATM adaption layer (AAL), which provides for call control across the ATM network and for segmentation and reassembly of frames from the data link layer. The current estimate for the amount of overhead needed per cell for AAL purposes is 4 octets, leaving 44 octets for user information. At the present time, end-to-end connections at the ATM level are expected to be connection oriented. As cells traverse the network, they are switched on a one-by-one basis, using information contained in the five ATM overhead octets to follow the virtual path established during the ATM call set-up. Typically, cells outbound on a common link are statistically multiplexed, and if buffers are full, cells are dropped. In addition, if one or more errors are found in a cell, then the cell is dropped. In the case of data transmission, a lost cell will result in an unusable frame unless the data is encoded to guard against cell loss prior to transmission. Coding might be provided by the AAL, for example. The trade- offs involved in coding and retransmission and their impact upon network throughput, delay and complexity are not well understood at the time of this writing. Part of the reason for this is that cell loss probability and FIGURE 72.8Asynchronous transfer mode layered architecture. ? 2000 by CRC Press LLC the types of traffic that are likely to use the network are not thoroughly understood. Resolution of these issues accounts for a significant portion of the research activity in computer communication networking at this time. The relevant American National Standards Institute and CCITT documents are frequently updated to include the results. Defining Terms Access line: A communication line that connects a user’s terminal equipment to a switching node. Adaption sublayer: Software that is added between two protocol layers to allow the upper layer to take advantage of the services offered by the lower layer in situations where the upper layer is not specifically designed to interface directly to the lower layer. Architecture: The set of protocols defining a computer communication network. Asynchronous transfer mode (ATM): A mode of communication in which communication takes place through the exchange of tiny units of information called cells. Broadband integrated services digital network (B-ISDN): A generic term that generally refers to the future network infrastructure that will provide ubiquitous availability of integrated voice, data, imagery, and video services. Carrier-sense multiple access: A random-access method of sharing a bus-type communications medium in which a potential user of the medium listens before beginning to transmit. Circuit switching: A method of communication in which a physical circuit is established between two terminating equipments before communication begins to take place. This is analogous to an ordinary phone call. Common-channel interoffice signaling: The use of a special network, dedicated to signaling, to establish a path through a communication network, which is dedicated to the transfer of user information. Computer communication network: Collection of applications hosted on different machines and intercon- nected by an infrastructure that provides intercommunications. Connection-oriented service: A mode of packet switching in which a call is established prior to any infor- mation exchange taking place. This is analogous to an ordinary phone call, except that no physical resources need to be allocated. Connectionless service: A mode of packet switching in which packets are exchanged without first establishing a connection. Conceptually, this is very close to message switching, except that if the destination node is not active, then the packet is lost. Entity: A software process that implements a part of a protocol in a computer communication network. Fast packet networks: Networks in which packets are transferred by switching at the frame layer rather than the packet layer. Such networks are sometimes called frame relay networks. At this time, it is becoming in vogue to think of frame relay as a service, rather than transmission, technology. Formal parameters: The parameters passed during the invocation of a service primitive; similar to the arguments passed in a subroutine call in a computer program. International Standards Organization reference model: A model, established by ISO, that organizes the functions required by a complete communication network into seven layers. Internet: A network formed by the interconnection of networks. Local-area network: A computer communication network spanning a limited geographic area, such as a building or college campus. Media-access control: A sublayer of the link layer protocol whose implementation is specific to the type of physical medium over which communication takes place and which controls access to that medium. Message switching: A service-oriented class of communication in which messages are exchanged among terminating equipments by traversing a set of switching nodes in a store-and-forward manner. This is analogous to an ordinary postal system. The destination terminal need not be active at the same time as the originator in order that the message exchange take place. Metropolitan-area network: A computer communication network spanning a limited geographic area, such as a city; sometimes features interconnection of LANs. Packet switching: A method of communication in which messages are exchanged between terminating equipments via the exchange of a sequence of fragments of the message called packets. ? 2000 by CRC Press LLC Protocol data unit (PDU): The unit of exchange of protocol information between entities. Typically, a PDU is analogous to a structure in C or a record in Pascal; the protocol is executed by processing a sequence of PDUs. Service primitive: The name of a procedure that provides a service; similar to the name of a subroutine or procedure in a scientific subroutine library. Switching node: A computer or computing equipment that provides access to networking services. Token bus: A method of sharing a bus-type communications medium that uses a token to schedule access to the medium. When a particular station has completed its use of the token, it broadcasts the token on the bus, and the station to which it is addressed takes control of the medium. Token ring: A method of sharing a ring-type communications medium that uses a token to schedule access to the medium. When a particular station has completed its use of the token, it transmits the token on the ring, and the station that is physically next on the ring takes control. Trunk: A communication line between two switching nodes. Wide-area network: A computer communication network spanning a broad geographic area, such as a state or country. Related Topics 71.3 Protonic Networks ? 72.1 B-ISDN References D. Bertsekas and R. Gallagher, Data Networks, Englewood Cliffs, N.J.: Prentice-Hall, 1987. A. Bhargava and M. G. Hluchyj, “Frame losses due to buffer overflow in fast packet networks,” Proc. IEEE INFOCOM ’90, San Francisco, 1990, pp. 132–139. U. Black, Computer Networks, Protocols and Standards, 2nd ed., Englewood Cliffs, N.J.: Prentice-Hall, 1994. U. Black, TCP/IP and Retated Protocols, New York: McGraw-Hill, 1992. J. L. Hammond and P. J. P. O’Reilly, Performance Analysis of Local Computer Networks, Reading, Mass.: Addison- Wesley, 1986. S. Haykin, Communication Systems, 3rd ed., New York: Wiley, 1994. H. J. Helgert, Integrated Services Digital Networks. Reading, Mass.: Addison-Wesley, 1991. M. Rose, The Open Book: A Practical Perspective on OSF; Englewood Cliffs; N.J.: Prentice-Hall, 1990. M. Schwartz, Telecommunications Networks: Protocols, Modeling and Analysis, Reading, Mass.: Addison-Wesley, 1987. J. D. Spragins, Telecommunications: Protocols and Design, Reading, Mass.: Addison-Wesley, 1991. W. Stallings, Handbook of Computer-Communications Standards: The Open Systems Interconnection (OSI) Model and OSI-Related Standards, New York: Macmillan, 1990a. W. Stallings, Handbook of Computer-Communications Standards: Local Network Standards, New York: Mac- millan, 1990b. W. Stallings, Handbook of Computer-Communications Standards: Department of Defense (DOD) Protocol Stan- dards, New York: Macmillan, 1990c. A. S. Tanenbaum, Computer Networks, 2nd ed., Englewood Cliffs, N.J.: Prentice-Hall, 1988. M. E. Woodword, Communication and Computer Networks, IEEE Press, 1993. Further Information There are many conferences and workshops that provide up-to-date coverage in the computer communications area. Among these are the IEEE INFOCOM and ACM SIGCOMM conferences and the IEEE Computer Communications Workshop, which specialize in computer communications and are held annually. In addition, IEEE GLOBCOM (annual), IEEE ICC (annual), IFIPS ICCC (biannual), and the International Telecommuni- cations Congress (biannual) regularly feature a substantial number of paper and panel sessions in networking. The ACM Communications Review, a quarterly, specializes in computer communications and often presents summaries of the latest standards activities. IEEE Network, a bimonthly, contains tutorial articles on all aspects ? 2000 by CRC Press LLC of computer communications and includes a regular column on books related to the discipline. Additionally, IEEE Communications and IEEE Computer, monthly magazines, frequently have articles on specific aspects of networking. For those who wish to be involved in the most up-to-date activities, there are many interest groups on the Internet, a worldwide TCP/IP-based network, that specialize in some aspect of networking. The User’s Directory of Computer Networks (Digital Press, T. L. LaQuey, Ed.) provides an excellent introduction to the activities surrounding internetworking and how to obtain timely information. 72.3 Local-Area Networks Joseph Bannister and Mario Gerla The local-area network (LAN) is a communication network that interconnects computers and computer-based devices, such as file servers, printers, and graphics terminals. The LAN is characterized as being contained completely within the premises of a single business entity—which almost always owns and operates the network—and this distinguishes the LAN from public-domain networks such as metropolitan- or wide-area networks. The LAN, then, is normally restricted to a few hundred stations (i.e., devices that attach directly to the LAN) that span a limited geographical area, so that no two connected stations are separated by a distance of more than a few kilometers. Moreover, the LAN can be distinguished from the computer or backplane bus, which interconnects components, boards, or devices that comprise a single computer. The LAN uses serial—rather than parallel—transmission, which also differentiates it from the computer bus. In contrast to today’s wide-area networks, information is transmitted over LANs at high speeds and with very low error rates. The LAN often employs fully broadcast media, or physical media that allow each station’s transmissions to be received by all other stations. Thus, a broadcast capability is often an integral feature of the LAN. Frequently, the LAN also provides for multicasting, a generalization of broadcasting in which a specified subset of stations receives a transmission. LANs are based on a variety of technologies that include twisted copper-wire pairs, coaxial cable, optical fibers, wireless infrared and radio for signal transport, as well as several integrated circuit families for trans- mitters, receivers, and the implementation of low-level protocols. The topology of a LAN refers to the physical layout of the transmission media and the logical arrangement of the stations on those media. Four topologies are commonly used in LANs: the bus, ring, star, and tree topologies, which are illustrated in Fig. 72.9. The LAN Service Model Within the scope of the well-known Open Systems Interconnection seven-layer reference model, the LAN occupies the two bottom layers, namely, the physical and data link layers, as shown in Fig. 72.10. The physical layer specifies the most primitive services of the LAN, e.g., media characteristics, signal formats, waveforms, signaling rates, timing, and mechanical aspects of connectors, etc. The data link layer uses the services of the physical layer to provide multiple access for stations sharing the media. Station or network management, which is shown as a vertical “layer” in Fig. 72.10, is responsible for maintaining a necessary level of performance, fault detection and recovery, and configuration and security functions. The Physical Layer Since the physical layer provides the most primitive services to LAN users, this layer is most closely associated with the implementation technology of the LAN. At a fundamental level the transmission media can be either electrical or optical waveguides. The physical media can be laid out as one of those topologies illustrated in Fig. 72.9. However, certain media are better suited to some topologies, e.g., the bus is frequently used with electrical but not with optical media, because there is high insertion loss associated with taps in the latter. Signaling is also a critical element of the physical layer. Baseband modulation, in which digital signals (0s and 1s) are transmitted as electrical or optical pulses, is common in LANs because of its simplicity. Modulation of carriers is less common but especially important when several independent channels are employed. Amplitude, ? 2000 by CRC Press LLC frequency, and phase modulation have been used in community antenna television (CATV) systems to support multichannel LANs. Coherent lightwave systems, although still largely experimental, are expected to increase in importance because they permit multiplexing a large number of channels over a single-mode optical fiber. Also of increasing importance is atmospheric propagation of electromagnetic signals. The growing demand for mobile communication and ubiquitous computing is driving the development of the personal communication network, which is to be based on code-division multiple access. The signaling rates and formats are also part of the physical layer specification. Electrical media generally use Manchester baseband encoding, which has a 50% duty cycle and operates at rates below 100 Mb/s. Optical media often use the so-called 4B/5B intensity-modulation encoding, which achieves 80% efficiency by repre- senting 27 distinct symbols (of which 16 are data and 11 are control symbols) by five bits in such a manner that four consecutive 0s (i.e., low-light power levels) should never occur. Similarly, 8B/6T encoding is sometimes used with electrical media to encode an octet of data as six ternary digits, achieving an efficiency of 75%. Connector and cable-plant technology is another critical element of the physical layer. Thorough character- ization of the transmission media is required if users are to interoperate with each other. The type of cable—e.g., shielded or unshielded twisted copper-wire pairs, coaxial cable, and single-mode or multimode optical fiber—must be specified. Furthermore, the connectors between stations and the cable plant are defined as part of the physical layer. Stations can attach via passive taps or can actively repeat signals; in the latter case a bypass switch is usually provided as the station’s interface to the cable plant. The Data Link Layer The data link layer is often divided into two sublayers, i.e., the media-access and logical-link control (MAC and LLC) sublayers, as shown in Fig. 72.10. The LLC sublayer [see Logical Link Control] uses the services of the MAC sublayer to provide to its user a connection-oriented service between stations that includes flow and error control or a connectionless service that does little more than multiplex upper-layer connections. The FIGURE 72.9 LAN topologies. ? 2000 by CRC Press LLC connection-oriented LLC protocol gives the service user the illusion of having a dedicated point-to-point link between a pair of communicating stations. The MAC sublayer specifies the media access protocol that stations use to share the media. In fully broadcast media no more than one station may transmit at a time, so the MAC sublayer manages exclusive access to the broadcast media. The ring topology is well suited to a token-passing MAC protocol, which gives transmission rights to the station holding the token. The token is represented by a special packet that is passed sequentially from station to station. When a station recognizes the token, it seizes it and begins transmitting buffered packets, or passes the token to the next station if it has no packet to transmit. To limit the time that a station can hold the token, the MAC protocol can implement one of several disciplines: ?One-shot service, in which the station releases the token when it has transmitted one packet ?Exhaustive service, in which the station releases the token when it has no more packets to transmit ?Gated service, in which the station releases the token when all packets that were buffered at token- acquisition time have been transmitted ?Token-timing service, in which the station releases the token at the expiration of a timer The IEEE 802.5 token ring standard specifies a token-timing service discipline that requires transmissions to be completed within a fixed time after the token is seized, but implementations sometimes use the simpler one- shot service discipline. The ANSI X3T9.5 fiber distributed data interface (FDDI) standard uses an adaptive token-timing service discipline that is intended to guarantee a minimum amount of (synchronous) bandwidth to each station. A variation of the token ring protocol is the token bus protocol, which allows the token to be passed in a specified order. In the token bus protocol, which is often used with the bus or tree topologies, a station broadcasts the token, specifying the successor station in an address field of the token packet. Although all stations receive the token, only the addressed successor station can seize it. A MAC scheme that is widely used with the bus topology is carrier-sense multiple access (CSMA). A contention protocol, CSMA operates by allowing any station to transmit a buffered packet if it senses that the bus is idle. If two stations are ready to transmit their packets at nearly the same time, they will both sense that the bus is idle and their transmissions will collide, i.e., the superimposed bits of the packets will be garbled. The propagation delay—or time it takes for the packet to travel from one station to the other—dictates the window of vulnerability for CSMA; the larger the window, the more collisions are likely. To overcome the problem of collisions, CSMA is often enhanced with collision detection (CSMA/CD) by enabling stations to monitor their transmissions for the garbled bits associated with collisions. When a collision is detected, the station aborts its transmission and reschedules it by backing off for a period of time. The binary exponential backoff algorithm specifies that the random backoff time is drawn uniformly from the interval between 0 and 2 n – 1 time units, where n is the number of times the packet has collided. FIGURE 72.10LANservice model. ? 2000 by CRC Press LLC A time-slotted bus maintains on the bus a continuous stream of short, fixed-length frames that are initially empty but can be filled with data as they pass stations with waiting packets. The distributed queue, dual bus (DQDB) local- and metropolitan-area network uses two-directional buses so that a station can reserve on its downstream bus a frame for its upstream-destined packets. In the star topology stations are homed into a central hub which can manage their access to the media. Active hubs physically control media access, while passive hubs merely broadcast incoming packets to specific output ports. Linear combiner/dividers based on lithium niobate technology allow incoming optical signals to be combined and distributed to output ports according to electronically programmed combining and dividing ratios. A common scheme is to use time-division multiplexing with the star topology. The hub can serve as the central controller, allocating time slots to individual stations, or reservations can be used in the manner of a satellite-based network. The Management Layer LAN-specific network management functions are referred to as station management. Station management covers five areas—configuration, performance, fault, accounting, and security management. Monitoring and controlling the LAN are essential elements of station management. By monitoring the media, stations maintain a record of important measurements, such as the number of a specific type of packet transmitted or received, the number of different kinds of errors, and the source addresses of received packets. Such measurements are made available to an application in the station or to a management center. Thus are applications able to monitor and collect, correlate, and act upon key LAN statistics. Likewise, designated applications are able to effect changes in the LAN by writing to specific variables within stations, which collectively comprise the so-called management information base. For instance, station management informs the MAC sublayer of its unique LAN address by writing the value to a special MAC register. Some management functions are distributed across the LAN and are implemented at a low level. To recover after the failure of a dual-fiber cable, stations automatically enter into a procedure to reconfigure around the failure and reestablish connectivity. Although such procedures can be viewed as station management, they are sometimes specified as part of the MAC sublayer, because they are so tightly integrated with media access. Other Features The basic features of media access are often augmented to provide specialized services and features. Specialized LAN Services LAN users have special communication requirements that must be supported by the physical and data link layers. In particular, the MAC sublayer is responsible for providing specialized services. Although all MAC sublayers support asynchronous traffic by providing for the simple, best-effort delivery of packets, some MAC sublayers also support other classes of traffic. To synchronous traffic, which requires a set amount of preallocated bandwidth, the properly designed MAC sublayer guarantees a maximum packet response time. The adaptively timed token-passing protocol of FDDI is capable of supporting synchronous traffic, i.e., at token-capture time the station has a fixed amount of time during which to transmit synchronous packets, and the token is guaranteed to return within a certain amount of time. Isochronous traffic, which requires a fixed amount of traffic to be periodically delivered, is also accommodated by some MAC sublayers. DQDB uses preallocated time slots to provide isochronous service. Priorities are also important in LANs. Therefore many LANs transmit queued packets in accordance with priorities assigned to the packets. Prioritization can be on a LAN-wide basis or merely within the station. Most LANs offer some method for prioritizing the transmission of packets. Reliability and Availability Being a shared resource, the LAN should have a high degree of reliability and availability. The media should not be a single point of failure, and no individual station should be able to prevent—maliciously or other- wise—the delivery of service to other stations. LANs are designed to withstand both transient and permanent failures of the media and stations. ? 2000 by CRC Press LLC Transmitted information is subject to short bursts of errors and must also be protected. The connection- oriented service at the LLC sublayer is intended to recover from errors—such as garbled, dropped, or out-of- sequence packets—by positively acknowledging packets and retransmitting packets not acknowledged within the timeout window. The MAC sublayer usually provides error-detecting codes that can recognize an error burst of several consecutive bits (a favorite is the 32-bit cyclic redundancy code, which is easily implemented as a linear feedback shift register). Errors can also be recognized at the physical layer when they cause code violations, e.g., the absence of transitions in the Manchester or 4B/5B codes. Some LANs even use error- correcting codes for protecting time-sensitive information. Other protection mechanisms are used to tolerate cable breaks and station malfunctions. The use of fully broadcast media makes a LAN vulnerable to media failure, since this effectively partitions the stations into noncommunicating groups. To cope with this problem, redundant cables are provided and a mechanism for reconfiguring from the bad to the good cable is built into the LAN protocols. A popular approach for the token ring can be seen in the counter-rotating dual-ring scheme, which is illustrated in Fig. 72.11. If a cable segment or an active station fails, the stations adjacent to the failure can reconfigure the ring by executing “wrap-around” operations. The new configuration uses the spare cable in conjunction with the original cable to form a new ring. Given the complexity of such a reconfiguration procedure, it is usually necessary for station management to coordinate the actions of the stations. Special mechanisms for adding and removing stations to and from the LAN might also be required. Since the physical addition or removal of a station can disrupt the transmission of data, protocols for reestablishing a lost token could also be necessary. FIGURE 72.11Reconfiguration of dual counter-rotating rings. ? 2000 by CRC Press LLC The Importance of LAN Standards LAN standards play a central role in promoting the goal of universal connectivity among a community of users. The standardization of communication services and protocols allows all conforming implementations to exchange information. Consequently, the importance of LAN standards has grown steadily. Currently, several LAN standards have been established to support the different communication requirements of users. The first LANs—developed in the early 1970s—were proprietary products meant to interconnect one vendor’s computer products. By 1980, however, Project 802 of the Institute of Electrical and Electronics Engineers (IEEE) had recognized the need for publicly disseminated LAN standards and eventually published a specification of the CSMA/CD protocol that any vendor may implement. Furthermore, the definition of the standard was sanctioned by companies that participated in the IEEE Working Group’s balloting process, so that the standard was viewed as an open, nonproprietary solution. The IEEE 802.3 Working Group chose a protocol that was based closely on the Ethernet LAN originally developed at Xerox by Robert Metcalfe and David Boggs. The IEEE Project 802 has broadened its scope to encompass other LAN standards. These include the following: ?802.3: The CSMA/CD protocols for baseband coaxial cable (10Base5 and 10Base2), unshielded twisted copper-wire pairs (10BaseT), broadband coaxial cable (10Broad36), and optical fiber (10BaseF) ?802.4: The token bus protocol for multichannel broadband coaxial cable ?802.5: The token ring protocol for shielded twisted copper-wire pairs ?802.6: The DQDB protocol for redundant optical fibers Other standards-making bodies, such as the American National Standards Institute (ANSI) and the Inter- national Organization for Standards/International Electrotechnical Committee (ISO/IEC) have developed or cross-adopted LAN standards. ANSI’s X3T9.5 committee is responsible for the FDDI LAN standard, a high- speed token ring that uses redundant optical fibers. Some of the important LAN standards and their charac- teristics are shown in Table 72.1. The trend is for vendors to market LAN products that conform to specific standards. However, proprietary networks have been successfully marketed and were instrumental in the development of LAN standards. Some of the better known proprietary-LAN product offerings were the Xerox Ethernet, Datapoint Arcnet, Network Systems Hyperchannel, Proteon Pronet, Sytek System 20, and AT&T DATAKIT. TABLE 72.1Characteristics of Standard LANs CSMA/CD Token Ring Token Bus FDDI DQDB Standard IEEE 802.3 IEEE 802.5 IEEE 802.4 ANS X3T9.5 IEEE 802.6 Topology Bus, tree, star Ring Tree Ring Pseudobus Media Coax, UTP, MMF STP Coax MMF, SMF SMF Encoding MC, FSK, AM/PSK 4B/5B, 8B/6T DMC FSK, AM/PSK 4B/5B 4B/5B Data rate 10 Mb/s, 100 Mb/s 4 Mb/s 1 Mb/s 100 Mb/s 34 Mb/s 16 Mb/s 5 Mb/s 45 Mb/s 10 Mb/s 140 Mb/s 155 Mb/s Features Priorities Priorities, ST, Priorities, ST, Priorities, IT, multichannel dual ring dual bus AM/PSK = amplitude modulation/phase-shift keying MC = Manchester coding ANS = American National Standard MMF = multimode fiber CSMA/CD = carrier-sense multiple-access with collision detection SMF = single-mode fiber DMC = differential Manchester coding ST = synchronous traffic DQDB = distributed queue, dual bus STP = shielded twisted pair FDDI = fiber distributed data interface UTP = unshielded twisted pair FSK = frequency-shift keying nB/mB = n-bit/m-bit IEEE = Institute of Electrical and Electronics Engineers nB/mT = n-bit/m-trit IT = isochronous traffic ? 2000 by CRC Press LLC Summary The LAN is the preferred method for connecting computers within a customer’s premises. A number of transmission media, topologies, data rates, and services are available to meet users’ needs. The services offered by the LAN are used to implement higher-layer protocols that are required by distributed computing systems. LANs will continue to grow more capable in the data and bit-error rates they achieve, the functionality they provide, and the number and geographical span of the stations they support. Defining Terms Media-access protocol: The protocol that permits one of a group of contending stations to access the media exclusively. Media-access protocols are generally based on token passing or carrier sense. Physical media: The communication channel over which signals are transmitted. Broadcast media, in which all stations receive each transmission, are primarily used in local-area networks. Common media are optical fibers, coaxial cable, twisted copper-wire pairs, and airwaves. Topology: The paths and switches of a local-area network that provide the physical interconnection among stations. The most common topologies are the bus, ring, tree, and star. Related Topics 72.2 Computer Communication Networks ? 75.3 Wireless Local-Area Networks for the 1990s References American National Standard for Information Systems—Fiber Distributed Data Interface (FDDI), ANSI Standards X3.139, X3.148, X3.166, X3.184. Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifica- tions, ANSI/IEEE Standard 802.3, ISO/IEC Standard 8802/3. Distributed Queue Dual Bus (DQDB) Metropolitan Area Network (MAN), Proposed IEEE Standard 802.6. Logical Link Control, ANSI/IEEE Standard 802.2, ISO/IEC Standard 8802/2. Token-Passing Bus Access Method and Physical Layer Specifications, ANSI/IEEE Standard 802.4, ISO/IEC Standard 8802/4. Token Ring Access Method and Physical Layer Specifications, ANSI/IEEE Standard 802.5, ISO/IEC Standard 8802/5. Further Information A popular, frequently updated textbook on LANs is W. Stallings, Local Networks, 3rd ed., New York: Macmillan, 1990. Leading journals that publish research articles on LANs include: ? IEEE Transactions on Communications ? IEEE Transactions on Networking ? Computer Networks and ISDN Systems ? IEEE Network Four annual conferences that cover the topic of LANs are: ? The IEEE INFOCOM Conference on Computer Communications ? The IEEE Conference on Local Computer Networks ? The ACM SIGCOMM Conference on Communications Architectures and Protocols ? The EFOC/LAN European Fibre Optic Communications and Local Area Networks Conference ? 2000 by CRC Press LLC 72.4 The Intelligent Network Richard B. Robrock II The term intelligent network refers to the concept of deploying centralized databases in the telecommunications network and querying those databases to provide a wide variety of network services such as 800 Service (toll- free service) and credit card calling. The first use of these centralized databases was in AT&T’s network in 1981 where they were used to facilitate the setup of telephone calls charged to a Calling Card. Today such databases are widely deployed throughout North America and support the handling of close to 100 billion telephone calls per year. The words intelligent network, when first used, had a relatively narrow definition, but that definition has broadened considerably with the introduction of the advanced intelligent network, the wireless intelligent network, and soon, the broadband intelligent network. The advanced intelligent network has introduced powerful service creation tools which have empowered network providers to create their own network services. The network providers, in turn, are beginning to broaden the participation in service creation by allowing their customers or third parties to use these tools to create services. The result has been a rapid growth in new network services. A History of Intelligence in the Network The first “intelligence” in the telephone network took the form of rows of human telephone operators, sitting side by side, plugging cords into jacks to facilitate the handling of calls. These operators established calls to far- away points, selected the best routes and provided billing information. They were also an information source—providing time or weather or perhaps disseminating the local news. Moreover, they had the opportunity to demonstrate a kind of heroism—gathering volunteers to save a house from fire, helping to catch a prowler, locating a lost child, and on and on. In the early years of telephony, the feats of the telephone operator were indeed legendary. In the 1920s, however, technology became available that allowed automatic switching of telephone calls through the use of sophisticated electromechanical switching systems. Initially, these switches served as an aid to operators; ultimately, they led to the replacement of operators. The combination of the rotary telephone dial and the electromechanical switch allowed customers to directly dial calls without the assistance of operators. This led to a reduction of human intelligence in the network. Another dramatic change took place in the telephone network in 1965; it was called software. It came with the marriage of the computer and the telephone switching system in the first stored-program control switch. With the introduction of switching software came a family of Custom Calling services (speed calling, call waiting, call forwarding, and three-way calling) for residential customers, and a robust set of Centrex features (station attendant, call transfer, abbreviated dialing, etc.) for business customers. The first software programs for these stored-program control switches contained approximately 100,000 lines of code; by 1990 some of these switching systems became enormously complex, containing 10 million lines of code and offering hundreds of different services to telephone users. During the 1980s, a new architectural concept was introduced; it came to be called the intelligent network. It allowed new telecommunications services to be introduced rapidly and in a ubiquitous and uniform fashion. Feature and service availability in the network ceased to be solely dependent upon the hardware and software in stored-program control switches. Rather some new intelligence was centralized in databases which were accessed using packet switching techniques. Most significantly, the intelligent network started to provide some of the capabilities that operators had made available in the early years of telephony. The remaining sections of this chapter describe the intelligent network, its characteristics, and its services. They also provide a description of the advanced intelligent network, which dramatically broadens the participation in the creation of new services. The Intelligent Network The intelligent network architecture is illustrated in Fig. 72.12; its primary elements are a switching system, a signaling network, a centralized database, and an operations support system which supports the database. The ? 2000 by CRC Press LLC architectural concept is a simple one. When a customer places a telephone call which requires special handling, such as a toll-free call (800 Service) or credit card call, that call is intercepted by the switching system which suspends call processing while it launches a query through a signaling network to a centralized database. The database, in turn, retrieves the necessary information to handle the call and returns that information through the signaling network to the switch so that the call can be completed. The role of the operations support system is to administer the appropriate network and customer information that resides in the database. It is conceivable that the database in this architecture could reside in the switching system, and the signaling network in this instance would not be required. However, that would magnify the task of administering the customer information, since that information would be contained in thousands of switches instead of dozens of centralized databases. In addition, even more importantly, there are two shortcomings associated with basing many of the potential new services in switches, rather than utilizing centralized databases to provide information for the switches. The first is a deployment problem. As of 1990 there were more than 15,000 switches in the United States, and a single switch can cost millions of dollars. To introduce a new service in local switches and to make it widely available generally requires some not-so-simple changes in those switches or, in some cases, replacement of certain switch types altogether. These switch modifications typically take years to implement and require a tremendous capital investment. As a result, ten years after introduction, Custom Calling services were available to fewer than 1% of the residential customers in the United States. A second problem with switch-based services has been that a single service sometimes behaves differently in different switch types. For example, the speed calling access patterns are different in various stored-program control switches. The public is not particularly sensitive to this fact, because speed calling is not associated with an individual but rather an individual’s station set. People live in a mobile society, however, and they want to have their services available from any station set and have them behave the same from any station set. FIGURE 72.12 Intelligent network architecture—telephone calls which require special handling are intercepted in a switching system which launches queries through a signaling network to a centralized database. (Source: R. B. Robrock II, “The intelligent network—Changing the face of telecommunications,” Proc. IEEE, vol. 79, no. 1, pp. 7–20, January 1991. ? 1991 IEEE.) FIGURE 72.13 Link arrangements in a CCS7 signaling network. (Source: R. B. Robrock II, “The intelligent net- work—Changing the face of telecommunications,” Proc. IEEE, vol. 79, no. 1, pp. 7–20, January 1991. ? 1991 IEEE.) ? 2000 by CRC Press LLC The intelligent network architecture has been the key to solving both the deployment problem and service uniformity problem associated with switch-based services. Services deployed using an intelligent network centralized database are immediately ubiquitous and uniform throughout a company’s serving area. Intelligent Network Systems In 1981, AT&T introduced into the Bell System a set of centralized databases called network control points; they supported two applications—the Billing Validation Application for Calling Card Service (credit card calling) and the INWATS database used to support 800 Service. Queries were launched to these databases through AT&T’s common-channel interoffice signaling (CCIS) network. In 1984, following the divestiture of the Regional Bell Operating Companies from AT&T, the regional companies began planning to deploy their own common-channel signaling (CCS) networks and their own centralized databases. They selected the signaling system 7 protocol for use in their signaling networks, called CCS7 networks, and they named their databases service control points (SCPs). The CCS7 Network A general architecture for a regional signaling network is shown in Fig. 72.13. The network is made up of signal transfer points (STPs), which are very reliable, high-capacity packet switches that route signaling messages between network access nodes such as switches and SCPs. To perform these routing functions, the STPs each possess a large routing database containing translation data. The CCS7 network in Fig. 72.13 contains both local STPs and regional STPs. The STPs are typically deployed in geographically separated pairs so that in the event of a natural disaster at one site, such as an earthquake, flood, or fire, the total traffic volume can be handled by the second site. Indeed, redundancy is provided at all key points so that no single failure can isolate a node. As illustrated in Fig. 72.13, the following link types have been designated: ?A-links connect an access node, such as a switching system or SCP, to both members of an STP pair. ?B-links interconnect two STP pairs forming a “quad” of four signaling links where each STP indepen- dently connects to each member of the other pair. ?C-links are the high-capacity connections between the geographically separated members of an STP pair. ?D-links connect one STP pair to a second STP pair at another level in the signaling hierarchy or to another carrier. ?E-links connect an access node to a remote STP pair in the signaling network and are rarely used. ?F-links directly interconnect two access nodes without the use of an STP; they are nonredundant. The CCS7 links normally function at 56 kb/s in North America while links operating at 64 kb/s are common in Europe. The CCS7 signaling network provides the underlying foundation for the intelligent network, and the regional telephone companies in the United States began wide-scale deployment of these networks in 1986; several large independent telephone companies and interexchange carriers (ICs) soon followed. They used these networks for both trunk signaling between switches as well as for direct signaling from a switch to a database. The Service Control Point The “brains” of the intelligent network is the SCP. It is an on-line, fault-tolerant, transaction-processing database which provides call handling information in response to network queries. The SCP deployed for 800 Service is a high-capacity system capable of handling more than 900 queries per second or 3 million per hour. It is a real-time system with a response time of less than one half second, and it is a high-availability system with a downtime of less than 3 minutes per year for a mated SCP pair. The SCP is also designed to accommodate growth, which means that processing power or memory can be added to an in-service system without inter- rupting service. In addition, it is designed to accommodate graceful retrofit, which means that a new software program can be loaded into an in-service SCP without disrupting service. ? 2000 by CRC Press LLC Data Base 800 Service SCPs have been deployed throughout the United States in support of the Data Base 800 Service mandated by the Federal Communications Commission. This service provides its subscribers with number portability so that a single 800 number can be used with different carriers. The Data Base 800 Service architecture is shown in Fig. 72.14. With this architecture, 800-number calls are routed from an end office to a service switching point (SSP) which launches queries through a CCS7 signaling network to the SCP. The SCP identifies the appropriate carrier, as specified by the 800 Service subscriber, and then, if appropriate, translates the 800 number to a plain old telephone (POTS) number. This information is subsequently returned to the SSP so that the call can be routed through the network by handing the call off to the appropriate carrier. This technology allows subscribers to select the carrier and the POTS number as a function of criteria such as time of day, day of week, percent allocation, and the location of the calling station. Thus the SCP provides two customer-specified routing information functions: a carrier identification function and an address translation function. The SCP 800 Service database is administered by a single national service management system (SMS). The SMS is an interactive operations support system that is used to process and update customer records. It is the interface between the customer and the SCP. It translates a language which is friendly to a customer into a language which is friendly to on-line, real-time databases. Along the way, it validates the customer input. Alternate Billing Services Alternate billing services (ABS) have also been implemented using the intelligent network architecture. Alternate billing is an umbrella title which includes Calling Card Service, collect calling, and bill-to-third-number calling. The network configuration supporting ABS is shown in Fig. 72.15. With this architecture, when a customer places a Calling Card call, the call is routed to an operator services system (OSS) which suspends call processing and launches a query through a CCS7 signaling network. The query is delivered to an SCP which contains the line information database (LIDB) application software. The LIDB application can provide routing information, such as identifying the customer-specified carrier which is to handle the call, as well as provide screening functions, such as the Calling Card validation used to authorize a call. The LIDB then returns the appropriate information to the OSS so that the call can be completed. The LIDBs are supported by the database administration system (DBAS), which is an operations support system that processes updates for Calling Card Service as well as other services. Multiple DBAS systems typically support each LIDB. During 1991, each of the Regional Bell Operating Companies and a number of large independent telephone companies interconnected their CCS7 networks, mostly through STP hubs, to create a national signaling network; it was a process called LIDB interconnect. When it was finished, it meant that a person carrying a FIGURE 72.14Data Base 800 Service—800-number calls are routed to an SSP which launches queries through a CCS7 network to an SCP containing the 800 database. In this example, the SCP translates the 800 Service number of 800-555- 5463 into the POTS number of 404-555-1000. (Source: R. B. Robrock II, “The intelligent network—Changing the face of telecommunications,” Proc. IEEE, vol. 79, no. 1. pp. 7–20, January 1991. ? 1991 IEEE.) ? 2000 by CRC Press LLC particular company’s Calling Card could, from anywhere in the United States, query the LIDB containing the associated Calling Card number. Although the LIDB was originally developed to support Calling Card Service, it has since found wide application in the telecommunications industry. For example, the LIDB is used to translate the telephone number of a calling party to a name as part of Calling Name Delivery service, or to convert that number to a nine-digit ZIP code as part of Single Number Service. The LIDB databases now contain more than a quarter of a billion customer records which are updated at a rate of more than a million changes per day. Although physically distributed, the LIDBs appear logically as a single database. They represent a national resource. Other Services For alternate billing services, the SCP is essentially designed to perform two functions: carrier identification and billing authorization. For 800 Service, the SCP provides carrier identification and address translation. These basic functions of authorization, address translation and carrier identification can be used again and again in many different ways. For example, the intelligent network has been used to support private virtual networks (PVNs). PVNs make use of the public telephone network but, by means of software control, appear to have the characteristics of private networks. A PVN serves a closed-user group, and a caller requires authorization to gain access to the network. This screening function on originating calls uses an authorization function. Second, a PVN may offer an abbreviated dialing plan, for example, four-digit dialing. In this instance, the SCP performs an address translation function, converting a four-digit number to a ten-digit POTS number. There may also be a customer-specified routing information function which involves selecting from a hierarchy of facilities; this can be accomplished through use of the SCP carrier identification function. The SCP in the intelligent network can support a vast number of services ranging from Calling Name Delivery service to messaging service. With Calling Name Delivery, a switch sends a query to the SCP with the ten-digit calling party number; the response is the calling party name which is then forwarded by the switch to a display unit attached to the called party station set. In support of messaging services, the address translation capability of the SCP can be used to translate a person’s telephone number to an electronic-mail address. As a result, the sender of electronic mail need only know a person’s telephone number. The Advanced Intelligent Network The intelligent network architecture discussed thus far is often referred to in the literature as Intelligent Network/1; this architecture has addressed the deployment problem and the service uniformity problem. The next phase in the evolution of this network has come to be called the advanced intelligent network (AIN), with the AIN standards defined by Bellcore. FIGURE 72.15 Alternate billing services—calls are routed to an OSS which launches queries through the CCS7 network to SCPs containing the LIDB application. (Source: R. B. Robrock II, “The intelligent network—Changing the face of telecommunications,” Proc. IEEE, vol. 79, no. 1. pp. 7–20, January 1991. ? 1991 IEEE.) ? 2000 by CRC Press LLC The concept of AIN is that new services can be developed and introduced into the network without requiring carriers to wait for switch generics to be upgraded. Some AIN applications introduce powerful service-creation capabilities which allow nonprogrammers to invoke basic functions offered in the network and stitch together those functions, as illustrated in Fig. 72.16, to constitute a new service. As a result, AIN promises to dramatically shorten the interval required to develop new services. Perhaps of greater significance, it promises to broaden the participation in service creation. In addition, it offers the opportunity to personalize or customize services. The silicon revolution has driven the cost of memory down to the point where it is economically viable to have enough memory in the network to store the service scripts or call processing scenarios that are unique to individuals. Many people think of the AIN as a collection of network elements, network systems and operations systems; this view might be called a technologist’s view. Perhaps a better representation is shown in Fig. 72.17; it shows a collection of people—people empowered to create services. Historically, the creation of new services provided by the telephone network has been the sole domain of the network element and network system suppliers. There is perhaps a good analogy with the automobile FIGURE 72.16Creating the service script or scenario for a call by stitching together functional blocks. (Source: R. B. Robrock II, “The intelligent network—Changing the face of telecommunications,” Proc. IEEE, vol. 79, no. 1. pp. 7–20, January 1991. ? 1991 IEEE.) FIGURE 72.17The advanced intelligent network—a business perspective. (Source: R. B. Robrock II, “Putting the Telephone User in the Driver’s Seat,” International Council for Computer Communication Conference on Intelligent Networks, pp. 144–150, May 1992.) ? 2000 by CRC Press LLC industry. A market study in the early 1900s predicted that 200,000 was the maximum number of cars that could ever be sold in a single year in the United States; the reasoning was that 200,000 was the maximum number of chauffeurs that could enter the workforce in a single year. In the telecommunications business, the network element and network system suppliers have been the chauffeurs of the network services business. The service-creation tools offered by the AIN, however, empower telephone company staff to create new services. Moreover, similar tools may well be used by the telecommunications staff of large corporations, or by third-party application providers or even by some segment of the telephone user population. As a result, we may see an explosion in the number of network services. The AIN introduces very powerful service-creation tools which are used to produce service-logic scripts (programs). In one arrangement, the service creation is done by assembling service-logic graphs from graphical icons that represent functional components of services. The completed graph is then validated with an expert system and tested off-line by executing every leg of the service-logic graph. At this point the service-logic program can be downloaded into the service control point so that it is ready for execution. To make use of the new service, it is then necessary to set “triggers” in the appropriate service switching point. These triggers can be set for both originating and terminating calls, and they represent events which, should they occur, indicate the need for the switch to launch a query to the SCP for information the switch needs to process the call. The AIN switch generics, which are presently deployed, support several triggers such as “immediate off hook” or “called address.” Future AIN switch generics are expected to support several dozen triggers. The first phase of the AIN, called AIN 0, became reality in late 1991 when friendly user trials began in two of the Regional Bell Operating Companies. AIN 0 evolved to AIN 0.1 and then AIN 0.2, with each new version of AIN containing additional triggers. Today over 100 AIN services are deployed in North America and the number is growing rapidly. The European Telecommunications Standards Institute (ETSI) has defined a European AIN standard referred to as Core INAP, and deployment of Core INAP systems in Europe began in 1996. The architectural concepts of AIN are now beginning to carry over into wireless networks as well as broadband networks. Although the standards in these domains are just being developed, the value added by the wireless intelligent network (WIN) and the broadband intelligent network (BIN) promises to surpass the value seen in the narrowband wireline world. Back to the Future The intelligent network, with its centralized databases, has offered a means to rapidly introduce new services in a ubiquitous fashion and with operational uniformity as seen by the end user. The advanced intelligent network has gone on to provide a service-independent architecture, and, with its powerful service-creation capabilities, has empowered nonprogrammers to participate in the development of new services. In many ways, as we go into the future, we are going back to a time when operators were the “human intelligence” in the network. The human intelligence was all but eliminated with the introduction of switching systems, but now the intelligent network is working to put the intelligence of the human operator back into the network. Defining Terms Common-channel signaling (CCS): A technique for routing signaling information through a packet-switched network. Database administration systems (DBAS): An operations support system that administers updates for the line information database. Line information database (LIDB): An application running on the service control point that contains infor- mation on telephone lines and Calling Cards. Service control point (SCP): An on-line, real-time, fault-tolerant, transaction-processing database which provides call-handling information in response to network queries. Service management system (SMS): An operations support system which administers customer records for the service control point. ? 2000 by CRC Press LLC Signal transfer point (STP): A packet switch found in the common-channel signaling network; it is used to route signaling messages between network access nodes such as switches and SCPs. Signaling system 7 (SS7): A communications protocol used in common-channel signaling networks. Related Topic 72.2 Computer Communication Networks References AT&T Bell Laboratories, “Common channel signaling,” The Bell System Tech. J., vol. 57, no. 2, pp. 221–477, February 1978. AT&T Bell Laboratories, “Stored program controlled network,” The Bell System Tech. J., vol. 61, no. 7, part 3, pp. 1573–1815, September 1982. Bell Communications Research, “Advanced intelligent network (AIN) 0.1 switch-service control point (SCP) application protocol interface generic requirements,” Bell Commun. Res. Technical Ref., TR-NWT-001285, Issue 1, August 1992. Bell Communications Research, “Advanced intelligent network (AIN) switch-service control point (SCP)/Adjunct interface generic requirements,” Bell Commun. Res., Generic Requirements, GR-1299- CORE, Issue 2, December 1994. European Telecommunications Standards Institute, “Intelligent network (IN): Intelligent network capability set 1 (CS1) core intelligent network applications protocol (INAP) part 1: Protocol specification,” Eur. Telecom. Stds. Inst., ETS 300 374-1, draft, May 1994. Globecom ’86: The Global Telecommunications Conference, Conference Record, vol. 3, pp. 1311– 1335, Decem- ber 1986. R.J. Hass and R.W. Humes, “Intelligent network/2: A network architecture concept for the 1990s,” International Switching Symposium, Conference Record, vol. 4, pp. 944–951, March 1987. R.B. Robrock, II, “The intelligent network—Changing the face of telecommunications,” Proc. IEEE, vol. 79, no. 1. pp. 7–20, January 1991. R.B. Robrock, II, “Putting the telephone user in the driver’s seat,” International Council for Computer Com- munication Intelligent Networks Conference, pp. 144–150, May 1992. R.B. Robrock, II, “The many faces of the LIDB data base,” International Conference on Communications, Conference Record, June 1992. Further Information The bimonthly magazine Bellcore Exchange has numerous articles on the intelligent network, particularly in the following issues: July/August 1986, November/December 1987, July/August 1988, and March/April 1989. Articles on AIN service creation appear in the January/February 1992 issue. Subscriptions or single copies are available from the Bellcore Exchange Circulation Manager, 60 New England Avenue, Piscataway, NJ 08854-4196. The monthly publication IEEE Communications Magazine contains numerous articles on the intelligent network. A special issue on the subject was published in January 1992. Copies are available from the IEEE Service Center, 445 Hoes Lane, Piscataway, NJ 08854-4150. The monthly publication The Bellcore Digest lists recent Bellcore publications. There are a series of technical advisories, technical requirements, and special reports that have been issued on the intelligent network. Copies are available by contacting Bellcore Customer Service Toll-Free 1-800-521-CORE (2673). The bimonthly publication The AT&T Technical Journal contains numerous articles on the intelligent net- work. The advanced intelligent network is the subject of a special issue: Summer 1991, vol. 70, nos. 3–4. Current or recent issues may be obtained from the AT&T Customer Information Center, P.O. Box 19901, Indianapolis, IN 46219. ? 2000 by CRC Press LLC