Biblio5

332 ENTERPRISE NETWORKS I

18. What is frame stripping?

19. How are collisions avoided using a token-passing scheme?

20. What is the function of a LAN repeater?

21. What are the four types of bridges covered in the text?

22. On what OSI layer do routers operate?

REFERENCES

1. IEEE Standard Dictionary of Electrical and Electronic Terms, 6th ed., IEEE Std. 100-1996, IEEE, New York, 1996.

2. Information Technology—Part 2, Logical Link Control, ANSI/ IEEE 802.2, 1994 ed., with Amd. 3, IEEE, New York, 1994.

3. Information Technology—Part 3: Carrier Sense Multiple Access with Collision Detection (CSMA/ CD) Access Method, ANSI/ IEEE Std. 802.3, 1996 ed., IEEE, New York, 1996.

4. Supplement to Carrier Sense Multiple Access with Collision Detection (CSMA/ CD) Access Method—100 Mbps Operation, Type 100BaseT, IEEE Std. 802.3u-1995, IEEE, New York, 1995.

5. Information Technology—Part 5: Token Ring Access Method and Physical Layer Specification, ANSI/ IEEE 802.5, 1995 ed., IEEE, New York, 1995.

6. Fiber Distributed Data Interface (FDDI)—Token Ring Physical Layer Medium Dependent, ANSI X3.166-1990, ANSI, New York, 1990.

7. Fiber Distributed Data Interface (FDDI) Physical Layer Protocol (PHY-2), ANSI X3.2311994, ANSI, New York, 1994.

8. Fiber Distributed Data Interface (FDDI): Token Ring Media Access Control—2, MAC-3, ANSI X.239-1994, ANSI, New York, 1994.

9. Internetworking Troubleshooting Seminar Presentation, Hewlett-Packard Co., Tempe, AZ, Jan. 1995.

10. ChipCom promotional material, ChipCom, Southboro, MA, 1995.

11. High-Speed Networking—Options and Implications, ChipCom, Southboro, MA, 1995.

334 ENTERPRISE NETWORKS II: WIDE AREA NETWORKS

12.1.1.1 Data Message Must Reach the Recipient(s). How is a data message routed such that the indicated recipient(s) receive the message? With conventional telephony, signaling carries out this function. It sets up a circuit (Chapter 7), maintains the connectivity throughout the call, and then takes the circuit down when one or both subscribers go on-hook. With data communications these same functions must be carried out. One form of data connectivity is called connection oriented, where indeed a circuit is set up, traffic is passed, and then the circuit is taken down. There is another form of data communications, where a frame is launched by the originator, and that frame must find its own way to the destination, similar to what happens using the postal service. A letter is dropped in an outgoing box, and it finds its way (with the help of the postal service) to the destination. We, the originators, are completely unaware of the letter’s routing; we do not really care. In the data world, this is called connectionless service. In the case of the postal service, the address on the envelope routed the letter.

For the data message case, it is the header that carries the routing information. In many situations, a data message may be made up of a number of frames. Each frame carries a field set aside for the recipient address. More often than not it is called the destination address. In many cases, a frame also carries the originator’s address. In either case, these are 8-bit fields, often extendable in increment of 8 bits.

Three key questions come to mind: (1) What part of a frame’s header can be easily identified as the originator and destination address fields? (2) What is the addressing capacity of an address field? Rephrased: how many distinct addresses can be accommodated in an 8-bit field? (3) Once a router, smart bridge, or data switch recognizes the boundaries of a destination address(es), how does it know how to route the frame?

1. What part of a frame’s header can be easily identified as the originator and destination address fields? A family of protocols governs the operation of a particular data network. Our interest here is in the network-layer and data-link layer protocols. These protocols carefully lay out the data-link frame and the network-layer frame (see Section 10.10). For example, in Section 10.7.2 we introduced synchronous transmission and the generic data-link layer frame (Figure 10.7). One thing a digital processor can do and do very well is count bits, and groups of 8 bits, which we call an octet (others call it a byte). A specific network utilizes a particular data-link layer protocol. Thus a processor knows a priori where field boundaries are, because it is designed to meet the requirements of a particular protocol. With some data-link protocols, however, there may be a variable length info field. It is obvious that if this is the case, the processor must be informed of the length of a particular info field. When this is so, info field length information is often found in a subfield inside the control field (see Figures 10.7 and 10.25). So to answer question 1, the digital processor knows a priori exactly which 8 (or 16) bits comprise the destination address field by simply counting down from the unique start-of-frame octet. It knows a priori because the router processor conforms to a particular protocol.

2. What is the addressing capacity of an address field? We will let each address consist of a distinct 8- or 16-bit binary sequence. For an 8-bit binary group, how many distinct 8-bit sequences are there? Remember, on developing code capacity (Section 10.4), that we can have 2n sequences where n is the number of bits in a particular sequence. In our sample case it is 8 bits. Thus, its addressing capacity is 28 or 256 distinct addresses. If extended addressing is employed, namely, where we use two octets for a destination address, often the least significant bit of the first octet is reserved to tell the receiver processor to expect the next octet also to be dedicated to destination

address. Now we have only 7 bits in the first octet left for addressing and all 8 bits in the second octet. Adding these two together, we get 15 bits for addressing. Thus, there are 215 or 32,768 distinct addresses, quite a respectable number.

3. How does a router or switching data node know how to route to a particular address? Simply by consulting a lookup table. Now this lookup table may have fixed routing entries, or entries that can be updated manually or routing entries that are updated dynamically. Here we must recognize three possible conditions:

1. A node/ router is added or dropped from the network;

2. New routing patterns are established; other routing patterns may be discontinued; and

3. Congestion, route/ node degradation and/ or failure may occur.

Read Section 12.3 on the TCP/ IP protocol family because IP has some very interesting means and methods to update routing/ lookup tables as well as finding routes to “unknown” destination addresses. Error detection and correction are discussed in Section 10.5.

12.1.1.2 There May Be an Issue of Urgency. There are several “families” of data messages that have limited or no real issue regarding urgency. For example, long accounting files, including payroll, may only require 24-hour or 48-hour delivery times. In this case, why not use the postal service, Federal Express, or UPS? On the other hand, credit card verification has high urgency requirements. A customer is waiting (probably impatiently) to have her/ his credit verified during the process of buying some item. Most transaction data messages are highly urgent.

One reason for implementing frame relay is its low latency. Let us call latency the time it takes to complete a data message transaction. There are four causes for an increase in transaction time:

1. Propagation delay;

2. The number of message exchanges required to complete a transaction (e.g., handshakes, circuit set-up, ARQ exchanges, and so on; X.25 is rich in such transactions);

3. Processing time and processing requirements (e.g., every effort has been made to reduce processing requirements in frame relay; there are virtually no message exchanges with frame relay); and

4. Secondarily, the quality of a circuit (if the circuit is noisy, many ARQ exchanges will occur increasing latency dramatically).

12.1.1.3 Recipient Must Be Prepared to Receive and “Understand” a Data Message. This is simply a question of compatibility. The data message receiver must be compatible with the far-end transmitter and intermediate nodes. We cannot have one transmitting the ASCII code and the companion receiver only able to receive EBCDIC. Such compatibility may extend through all seven OSI layers. For example, frame relay is based only on OSI layers 1 and 2. It is the responsibility of the frame relay user to provide the necessary compatibility of the upper OSI layers.

Now that the readers understand the three major points discussed above, we proceed with the following sections dealing with the most well-known generic WAN protocols.

336 ENTERPRISE NETWORKS II: WIDE AREA NETWORKS

Figure 12.1 X.25 packet communications operates with the public switched data network (PDN). CPU c central processing unit or mainframe computer.

12.2 PACKET DATA COMMUNICATIONS BASED ON CCITT REC. X.25

The concept of a packet-switched network is based on the idea that the network switching nodes will have multiple choices for routing of data packets. If a particular route becomes congested or has degraded operation, a node can send a packet on another route, and if that route becomes congested, possibly a third route will be available to forward the packet to its destination.

At a data source, a file is segmented into comparatively short data packets, each of the same length and each with its own header and trailer. As we mentioned, these packets may take diverse routes through various nodes to their destination. The destination node is responsible for data message reassembly in its proper order.

12.2.1 Introduction to CCITT Rec. X.25

The stated purpose of CCITT Rec. X.25 (Ref. 1) is to define an interface between the DTE and the DCE at the first three OSI layers. Ideally the DCE resides at the local data switching exchange and the DTE is located on customer premises. In addition X.25 defines the procedures necessary for accessing a packet-switched public data network (PDN). Figure 12.1 shows the X.25 concept of accessing the PDN. An example of the PDN is ISDN described in Section 12.4.

Data terminals defined by X.25 operate in a synchronous full-duplex mode with data rates of 2400, 4800, 9600, 14,400, 28,800, and 33,600 bps; 48, 64, 128, 192, 256, 384, 512, 1024, 1536, and 1920 kbps.

Figure 12.2 X.25 relationship with the OSI reference model.

12.2.2 X.25 Architecture and Its Relationship to OSI

X.25 spans the lowest three layers of the OSI reference model, as illustrated in Figure 12.2. It can be seen in the figure that X.25 is compatible with OSI up to the network layer. In this context there are differences at the network/ transport layer boundary. CCITT leans toward the view that the network and transport layer services are identical and that these are provided by X.25 virtual circuits.

12.2.2.1 User Terminal Relationship to the PDN. CCITT Rec. X.25 calls the user terminal the DTE, and the DCE resides at the related PDN node. The entire recommendation deals with this DTE–DCE interface, not just the physical layer interface. For instance, a node (DCE) may connect to a related user (DTE) with one digital link, which is covered by SLP (single-link procedure) or several links covered by MLP (multilink procedure). Multiple links from a node to a DTE are usually multiplexed on one transmission facility.

The user (DTE) to user (DTE) connectivity through the PDN based on OSI is shown in Figure 12.3. A three-node connection is illustrated in this example. Of course, OSI layers 1–3 are Rec. X.25 specific. Note that the protocol peers for these lower three layers are located in the PDN nodes and not in the distant DTE. Further, the Rec. X.25 protocol operates only at the interface between the DTE and its related PDN node and does not govern internodal network procedures.

12.2.2.2 Three Layers of X.25

12.2.2.2.1 Physical Layer. The physical layer is layer 1, where the requirements are defined for the functional, mechanical, procedural, and electrical interfaces between the DTE and DCE. CCITT Rec. X.21 or X.21 bis is the applicable standard for this interface. X.21 bis is similar to EIA-232 (See Section 10.8). CCITT Rec. X.21 specifies a 15-pin DTE–DCE interface connector. The electrical characteristics for this interface are the

338 ENTERPRISE NETWORKS II: WIDE AREA NETWORKS

Figure 12.3 X.25 user (DTE) connects through the PDN to a distant user (DTE).

same as CCITT Recs. V.10 and V.11, depending on whether electrically balanced or unbalanced operation is desired. These two recommendations have some similarity to the electrical portion of EIA-232.

12.2.2.2.2 CCITT X.25 Link Layer. The link layer of X.25 uses the LAPB protocol. LAPB is fully compatible with HDLC link-layer access protocol (see Section 10.10.3). The information field in the LAPB frame carries the user data, in this case the layer 3 packet.

LAPB provides several options for link operation. These include two versions of the control field: standard and extended. It also supports multilink procedures. MLP allows a group of links to be managed as a single transmission facility. It carries out the function of resequencing packets in the proper order at the desired destination. When MLP is implemented, an MLP control field of two octets in length is inserted as the first 16 bits of the information field. This field contains a multilink sequence number and four control bits (see Figure 12.4).

12.2.2.2.3 Datagrams, Virtual Circuits, and Logical Connections. There are three approaches used with X.25 operation to manage the transfer and routing of packet streams: datagrams, virtual connections (VCs), and permanent virtual connections (PVCs). Datagram service uses optimal routing on a packet-by-packet basis, usually over diverse routes. In the virtual circuit approach, there are two operational “modes”: virtual connection and permanent virtual connection. These two are analogous to a dial-up telephone connection and a leased line connection, respectively. With the virtual connection a logical connection is established before any packets are sent. The packet originator sends a call request to its serving node, which sets up a route in advance to the desired destination. All packets of a particular message traverse this route, and each packet of the message contains a virtual circuit identifier (logical channel number) and the packet data. At any one time each station can have more than one virtual circuit to any other station and can have virtual circuits to more than one station. With virtual circuits routing decisions are made in advance. With the datagram approach ad hoc decisions are made for each packet at each node. There is no call-setup phase with datagrams as there is with virtual connections. Virtual connections are advantageous for high community- of-interest connectivities, datagram service for low community-of-interest relations.

Datagram service is more reliable because traffic can be alternately routed around

Figure 12.4 Basic X.25 frame structure. Note how the X.25 packet is embedded in the LAPB frame. For the extended LAPB structure, the control field will have 16 bits. There is a basic packet structure and an extended packet structure. For the basic data packet, XX c 01. For an extended data packet, XX c 10. Q c qualifier bit; D is the delivery confirmation bit.

network congestion points. Virtual circuits are fixed-routed for a particular call. Callsetup time at each node is eliminated on a packet basis as with the virtual connection technique. X.25 also allows the possibility of setting up permanent virtual connections and is network assigned. This latter alternative is economically viable only for very- high-traffic relations; otherwise these permanently assigned logical channels will have long dormant periods.

12.2.2.3 X.25 Frame Structure: Layer 3, the Packet Layer. The basic data-link layer (LAPB) frame structure is given in Figure 12.4. Its similarity to the HDLC frame structure is apparent. For the X.25 case the packet is embedded in the LAPB information field, as mentioned. When applicable, the other part of the information field contains the MLP, which is appended in front of the X.25 packet and is the first subfield in the information (I) field. In fact, the MLP is not part of the actual packet and is governed by the layer 2 LAPB protocol.1

12.2.2.3.1 Structure Common to All Packets. Table 12.1 shows 17 packet types involved in X.25. Every packet transferred across the X.25 DTE–DCE interface consists of at least three octets.2 These three octets contain a general format identifier, a logical channel identifier, and a packet type identifier. Other fields are appended as required, as illustrated in Figure 12.4.

1LAPB is often called “link access protocol B-channel.” B-channel is ISDN nomenclature for a 64-kbps channel designated to carry revenue-bearing traffic.

2An octet is an 8-bit sequence. There is a growing tendency to use octet rather than byte when describing such a sequence. It removes ambiguity on the definition of byte.

a Not necessarily available on every network.

Note: A bit that is indicated as × may be set to either 0 or 1.

Source: ITU-T Rec. X.25, Table 5-2/ X.25, p. 52 (Ref. 1).

Now consider the general format identifier field in Figure 12.4. This is a 4-bit sequence. Bit 8, counting from right to left in Figure 12.4, is the qualifier bit (Q) found only in data packets. In the call-setup and clearing packets it is the A-bit, and in all other packets it is set to 0. The D-bit, when set to 1, specifies end-to-end delivery confirmation. This confirmation is provided through the packet receive number [P(R)]. When set to 01, the XX bits indicate a basic packet and when set to 10, an extended packet is indicated. The extension number involves sequence number lengths. (i.e., a longer sequence number can be accommodated).

Logical channel assignment is shown in Figure 12.5. The logical channel group and logical channel number subfields identify logical channels with the capability of identifying up to 4096 channels (212). This permits the DTE to establish up to 4095 simul-