Biblio5

302ENTERPRISE NETWORKS I

•The incorporation of such devices as repeaters, bridges, routers and switching hubs.

The preponderance of LANs operate without error correction with BERs specified in the range of 1 × 10−8 to 1 × 10−12 or better.

The most common application of a LAN is to interconnect data terminals and other processing resources, where all the devices reside in a single building or complex of buildings, and usually these resources have a common owner. A LAN permits effective cost sharing of high-value data-processing equipment, such as mass storage media, mainframe computers or minicomputers, and high-speed printers. Resource sharing is probably equally as important where a LAN serves as the access vehicle for an intranet. Resource sharing in this context means that each LAN user has access to all other users’ files and other data resources.

The interconnection of LANs in the local area (e.g., the floor of a building) with a high-speed backbone (e.g., between building floors) is very prevalent. LANs may connect through wide area networks (WANs) to other distant LANs. This is frame relay’s principal application. The interface to the WAN may be via a smart bridge or a router.

There are two generic transmission techniques utilized by LANs: baseband and broadband. Baseband transmission can be defined as the direct application of the baseband signal to the transmission medium. Broadband transmission, in this context, is where the baseband signal from the data device is translated in frequency to a particular frequency slot in the RF spectrum. Broadband transmission requires a modem to carry out the translation. Baseband transmission may require some sort of signal-conditioning device. With broadband LAN transmission we usually think of simultaneous multiple RF carriers that are separated in the frequency domain. Present broadband technology comes from the cable television (CATV) industry.

The discussion in this section will essentially cover baseband LANs. An important aspect is that only one user at a time may access the LAN.

11.3 LAN TOPOLOGIES

There are three types of basic LAN topology: bus, ring, and star.1 These are illustrated in Figure 11.1 along with the tree network, (Figure 11.1d ), which is a simple derivative of the conventional bus topology.

A bus is a length of transmission medium which users tap into, as shown in Figure 11.1a. Originally the medium was coaxial cable. Today coaxial cable is being phased out in favor of unshielded twisted pair (UTP) or fiber-optic cable.

A ring is simply a bus that is folded back onto itself. A ring topology is illustrated in Figure 11.1b. User traffic flows in one direction around the ring. In another approach (FDDI), which is discussed later in Section 11.6.4, a second ring is added where the traffic flow is in the opposite direction. Such a dual counterrotating ring concept improves reliability in case of a failed station or a cut in the ring.

A star network is illustrated in Figure 11.1c. At the center of the star is a switching device. This could be a switching hub. Users can be paired, two at a time, three at a time, or all at a time, segmented into temporary families of users, depending on the configuration of the switch at that moment in time. Such a concept lends itself particu-

1Topology means the logical and/ or physical arrangement of stations on a network. In other words, topology tells how these assets are connected together.

must have sufficient signal level and S/ N such that copied signals have a BER in the range of 1 × 10−8 to 1 × 10−12. If the medium is fairly long in extension and there are many accesses, the signal level must be high for a transmitting access to reach its most distant destination. The medium is lossy, particularly at the higher bit rates, and each access tap has an insertion loss. This leads to very high signal levels. These may be rich in harmonics and spurious emissions, degrading bit error rate. On the other hand, with insufficient level, the S/ N ratio degrades, which will also degrade error performance. A good level balance must be achieved for all users. Every multipoint connectivity must be examined. The number of multipoint connectivities can be expressed by n (n − 1), where n is the number of accesses. If, on a particular LAN, 100 accesses are planned, there are 9900 possible connectivities to be analyzed to carry out signal level balance. One way to simplify the job is to segment the network, placing a regenerative repeater (or bridge) at each boundary. This reduces the signal balance job to realizable proportions and ensures that a clean signal of proper level is available at each access tap. For baseband LANs, 50-Q coaxial cable is favored over the more common 75-Q cable. The lower-impedance cable is less prone to signal reflections from access taps and provides better protection against low-frequency interference.

11.5 OVERVIEW OF ANSI/ IEEE LAN PROTOCOLS

11.5.1 Introduction

Many of the widely used LAN protocols have been developed in North America through the offices of the Institute of Electrical and Electronic Engineers (IEEE). The American National Standards Institute (ANSI) has subsequently accepted and incorporated these standards, and they now bear the ANSI imprimatur.

The IEEE develops LAN standards in the IEEE 802 family of committees. Of interest in this chapter are the following IEEE committees, each with published standards, which continue to evolve:

802.1 High-Level Interface;

802.2 Logical Link Control (LCC);

802.3 CSMA/ CD Networks;

802.4 Token Bus Networks;

802.5 Token Ring Networks;

802.6 Metropolitan Area Networks (MANs); and

802.7 Broadband Technical Advisory Group.

The Fiber Distributed Data Interface (FDDI) standard, which is addressed in Section 11.6.4, was developed and is maintained by ANSI.

11.5.2 How LAN Protocols Relate to OSI

LAN protocols utilize only OSI layers 1 and 2, the physical and data-link layers, respectively. The data-link layer is split into two sublayers: logical link control (LLC) and medium access control (MAC). These relationships are shown in Figure 11.2. The principal functions of OSI layer 3, namely, switching, relaying, and network end-to-end control, are not necessary in this simple, closed network. Remaining layer 3 functions

306 ENTERPRISE NETWORKS I

Figure 11.2 LAN 802 architecture related to OSI.

that are necessary are incorporated in layer 2. The two layer-2 sublayers (LLC and MAC) carry out four functions:

1. Provide one or more service access points (SAPs). A SAP is a logical interface between two adjacent layers.

2. Before transmission, assemble data into a frame with address and error-detection fields.

3. On reception, disassemble the frame and perform address recognition and error detection.

4. Manage communications over the link.

The first function and those related to it are performed by the LLC sublayer. The last three functions are handled by the MAC sublayer.

In the following subsections we will describe four common IEEE and ANSI standardized protocols. Logical link control (LLC) is common to all four. They differ in the medium access control (MAC) protocol.

A station on a LAN may have multiple users; oftentimes these are just processes, such as processes on a host computer. These processes may wish to pass traffic to another LAN station that may have more than one “user” in residence. We will find that LLC produces a protocol data unit (PDU) with its own source and destination address. The source address, in this case, is the address of the originating user. The destination address is the address of a user in residence at a LAN station. Such a user is connected through a service access point (SAP) at the upper boundary of the LLC layer. The resulting LLC PDU is then embedded in the information field of a MAC frame. This is shown in Figure 11.3. The MAC frame also has source and destination addresses. These direct traffic to a particular LAN station or stations.

11.5.3 Logical Link Control (LLC)

The LLC provides services to the upper layers at a LAN station. It provides two forms of services for its users:

Figure 11.3 A user passes traffic to an LLC where encapsulation takes place forming an LLC PDU. The LLC PDU is embedded in a MAC frame info(rmation) field. The resulting MAC frame is passed to the physical layer, which transmits the traffic on the LAN.

1. Unacknowledged connectionless service; and

2. Connection mode services.

Some brief comments are required to clarify the functions and limitations of each service. With unacknowledged connectionless service a single service access initiates the transmission of a data unit to the LLC, the service provider. From the viewpoint of the LLC, previous and subsequent data units are unrelated to the present unit. There is no guarantee by the service provider of the delivery of the data unit to its intended user, nor is the sender informed if the delivery attempt fails. Furthermore, there is no guarantee of ordered delivery. This type of service supports point-to-point, multipoint, and broadcast modes of operation.

As we might imagine with connection mode service, a logical connection is established between two LLC users. During the data-transfer phase of the connection, the service provider at each end of the connection keeps track of the data units transmitted and received. The LLC guarantees that all data will be delivered and that the delivery to the intended user will be ordered (e.g., in the sequence as presented to the source LLC for transmission). When there is a failure to deliver, it is reported to the sender.

IEEE (Ref. 2) defines the LLC as that part of a data station that supports the LLC functions of one or more logical links. The LLC generates command PDUs and response PDUs for transmission and interprets received command PDUs and response PDUs. Specific responsibilities assigned to the LLC include:

•Initiation of control signal interchange;

•Interpretation of received command PDUs and generation of appropriate response PDUs;

•Organization of data flow; and

•Actions regarding error-control and error-recovery functions in the LLC sublayer.

As shown in Figure 11.3, the LLC accepts higher level user data and encapsulates it, forming an LLC PDU. The resulting LLC frame is embedded into the MAC user field for transmission.

The LLC is another derivative of HDLC, which was discussed in Section 10.10.3. It is based on the balanced mode of that link-layer protocol with similar formats and functions. This is particularly true when operating in the connection mode.

Figure 11.6 LLC PDU control field formats (Ref. 2).

However, only the last seven of these bits are used for the actual address. The first bit in the DSAP indicates whether the address is an individual address or a group address (i.e., addressed to more than one SAP). The first bit in the SSAP is the C/ R bit, which indicates whether a frame is a command frame or a response frame. The control field is briefly described in Section 11.5.3.2.

11.5.3.2 LLC Control Field and Its Operation. The LLC control field is illustrated in Figure 11.6. It is 16 bits long for formats that include sequence numbering and 8 bits long for formats that do not. The three formats described for the control field are used to perform numbered information transfer (I-frame), unnumbered control (S-frames), and unnumbered information transfer (U-frames) functions. These functions are described in Section 10.10.3.

11.6 LAN ACCESS PROTOCOLS

11.6.1 Introduction

In this context a protocol includes a means of permitting all users to access a LAN fairly and equitably. Access can be random or controlled. The random access schemes to be discussed include CSMA (carrier sense multiple access) and CSMA/ CD, where CD stands for collision detection. The controlled access schemes that are described are token bus and token ring. It should be kept in mind that users accessing the network are unpredictable, and the transmission capacity of the LAN should be allocated in a dynamic fashion in response to those needs.

11.6.2 CSMA and CSMA/ CD Access Techniques

Carrier sense multiple access (CSMA) is a LAN access technique that some simplistically call “listen before transmit.” This “listen before transmit” idea gives insight into

310 ENTERPRISE NETWORKS I

the control mechanism. If user 2 is transmitting, user 1 and all others hear that the medium is occupied and refrain from using it. In actuality, when an access with traffic senses that the medium is busy, it backs off for a period of time and tries again. How does one control that period of time? There are three methods of control commonly used. These methods are called persistence algorithms and are outlined briefly below:

•Nonpersistent. The accessing station backs off a random period of time and then reattempts access.

•1-Persistent. The station continues to sense the medium until it is idle and then proceeds to send its traffic.

•p -Persistent. The accessing station continues to sense the medium until it is idle, then transmits with some preassigned probability p. Otherwise it backs off a fixed amount of time, then transmits with a probability p or continues to back off with a probability of (1 − p).

The algorithm selected depends on the desired efficiency of the medium usage and the complexity of the algorithm and resulting impact on firmware and software. With the nonpersistent algorithm, collisions are effectively avoided because the two stations attempting to access the medium will back off, most probably with different time intervals. The result is wasted idle time following each transmission. The 1-persistent algorithm is more efficient by allowing one station to transmit immediately after another transmission. However, if more than two stations are competing for access, collision is virtually assured.

The p -persistent algorithm lies between the other two and is a compromise, attempting to minimize collisions and idle time.

It should also be noted that with CSMA, after a station transmits a message, it must wait for an acknowledgment from the destination. Here we must take into account the round-trip delay (2 × propagation time) and the fact that the acknowledging station must also contend for medium access. Another important point is that collisions can occur only when more than one user begins transmitting within the period of propagation time. Thus CSMA is an effective access protocol for packet transmission systems where the packet transmission time is much longer than the propagation time.

The inefficiency of CSMA arises from the fact that collisions are not detected until the transmissions from the two offenders have been completed. With CSMA/ CD, which has collision detection, a collision can be recognized early in the transmission period and the transmissions can be aborted. As a result, channel time is saved and overall available channel utilization capability is increased.

CSMA/ CD is sometimes called “listen while transmitting.” It must be remembered that collisions can occur at any period during channel occupancy, and this includes the total propagation time from source to destination. Even at multimegabit data rates, propagation time is not instantaneous; it remains constant for a particular medium, no matter what the bit rate is. With CSMA the entire channel is wasted. With CSMA/ CD one offending station stops transmitting as soon as it detects the second offending station’s signal. It can do this because all accesses listen while transmitting.

11.6.2.1 CSMA/ CD Description. Carrier sense multiple access with collision detection (CSMA/ CD) is defined by ISO/ IEC 8802-3 international standard and by the reference ANSI/ IEEE Std 802.3 (Refs. 3 and 4). Some refer to this as Ethernet, which was initiated by Xerox Corporation, Digital Equipment Corporation, and Intel. The version