A common realization of themultipointtopology is a network where all nodes physically connect to (and logically share) a common broadcast medium. Figure 4-3 shows the multipoint topology, where Nodes A through F communicate via a shared physical me-dium. Sometimes the shared medium is also called a common bus. Most local area net-works (LANs) utilize a broadcast (or multipoint) topology. Indeed, the IEEE 802.4 Token Bus, the IEEE 802.3 Ethernet, and the IEEE 802.6 Distributed Queue Dual Bus (DQDB) protocols define different means of logically sharing access to the common physical
me-42
ATM & MPLS Theory & Application: Foundations of Multi-Service NetworkingFigure 4-2. Point-to-point topology examples
Figure 4-3. Shared medium (common bus) multipoint (broadcast) topology
electromagnetic signal propagation.
A multidropanalog line is commonly used for legacy SNA Synchronous Data Link Control (SDLC) loop access, further described in Chapter 7. In this example, an analog signal is broadcast from a master station (usually a mainframe front end processor) to all slave stations. In the return direction, the slaves share the common broadcast medium of the multidrop line. The SNA SDLC polling protocol involves the host polling the slave stations in a round-robin manner, thus preventing any two slaves from transmitting at the same time. (See References [Cypser 78] and [Ranade 89] for more information on the SNA SDLC protocol.)
Other networks, notably the Ethernet protocol, also work on a broadcast medium but don’t provide for orderly coordination of transmissions as the SNA SDLC loop does.
Instead, these protocols empower stations to transmit whenever they need to as long as another station isn’t already sending data. When a collision does occur, a distributed al-gorithm uses the bandwidth at approximately 50 percent efficiency. Chapter 9 covers Ethernet and related local area networking protocols.
Figure 4-4 illustrates other conceptual examples of the multipoint topology. Another commonly used multipoint topology is that of broadcast, or multipoint-to-multipoint, which is the case where many other nodes receive one sender’s data. Yet another example
Figure 4-4. Conceptual illustration of multipoint topologies
is that of “incast,” or multipoint-to-point, where multiple senders’ signals are received at one destination—as in a slave-to-master direction. Some of the foundational work in MPLS embraced the multipoint-to-point concept with the aim of reducing the number of connection points within a network. In this conceptual illustration, note that the multipoint-to-multipoint (i.e., shared medium, or multicast) topology is effectively a full mesh of multipoint-to-point connections between each of the four nodes. The figure also illustrates emulation of a point-to-multipoint topology via multiple point-to-point links for comparison purposes.
Star
Thestartopology developed during the era when mainframes centrally controlled most computer communications. The voice-switched world also employs a star topology when multiple remote switching nodes, each serving hundreds to even thousands of tele-phone subscribers, home in on a large central switch. A variation on this topology is a dual star, which achieves enhanced resilience. This type of network radiates in a star-like fashion from the central switch through the remote switches to user devices. The central node performs the communication switching and multiplexing functions in the star to-pology. Nodes communicate with each other through point-to-point or multipoint links radiating from the central node. The difference between this topology and the multipoint topology is that the central node provides only point-to-point connections between any edge node, on either a physical or logically switched basis.
Figure 4-5 shows a star topology, where Node A serves as the center of the star and Nodes B through E communicate via connections switched to and through the central Node A. An example of a star topology is many remote terminal locations, or clients, accessing a
44
ATM & MPLS Theory & Application: Foundations of Multi-Service NetworkingFigure 4-5. Illustration of a network with a star topology
called a “hub and spoke” topology. The central hub may logically organize the physical star as a logical bus or ring, as is commonly done in LAN wiring hubs. A key benefit of the physical star topology is superior network management of the physical interfaces. For ex-ample, if a single interface fails in a physical star topology, then the management system can readily disable it without affecting any other stations. Conversely, in a broadcast to-pology, a single defective switch can take down the entire shared-medium network. As we shall see, many wide area networks also have a star topology, driven by the cli-ent/server computing paradigms.
Ring
Theringtopology utilizes a shared transmission medium that forms a closed loop. Such networks utilize protocols to share the medium and prevent information from circulating around the closed physical transmission circuit indefinitely. A ring is established, and each device passes information in one direction around the ring.
Figure 4-6 shows a ring network where in step1, Node A passes information ad-dressed around the ring through Node D in step2. Node C removes this frame from the ring and then returns a confirmation addressed to Node A in step3 via Node B, at which point Node A removes this data from the ring in step 4. Note that actual implementations of ring structures for LAN protocols use a more complicated protocol than that described here. Rings reuse capacity in this example because the destination removes the information from the ring so that other stations can utilize the ring bandwidth. Examples of the ring to-pology protocols are the IEEE 802.5 Token Ring and the Fiber Distributed Data Interface
Figure 4-6. Ring or loop topology
(FDDI). Although the ring topology looks like a special case of a mesh network, it differs because of the switching action performed at each node. SONET protection rings also use the ring topology, and they are also distinguished from a mesh by the difference in nodal switching action from that of a mesh of circuit switches.
Mesh
Many switched, bridged, and routed networks employ some form of mesh architecture.
Mesh networks have many nodes, which are connected by multiple links. If each node is directly connected to every other node, then the network is fully meshed; otherwise, the network is only partially meshed. Figure 4-7 shows a partial mesh network where Nodes B, C, D, E, F, and G have a high degree of connectivity by virtue of having at least three links to any other node, while Nodes A and H have only two links to other nodes. Note that Nodes C and G have four links. The number of links connected to a node is that node’s degree (of connectivity). For example, Node C has degree 4, while node H has degree 2.
Figure 4-8 shows afull meshnetwork where each node has a link to every other node.
Almost every major computer and data communications network uses a partial mesh to-pology to give alternate routes for backup and traffic loads. Few use a full mesh toto-pology, primarily because of cost and/or complexity factors associated with having a large num-ber of physical and/or logical links. This is because a full meshN-node network has N(N– 1) / 2 links, which is on the order ofN2for large values ofN.ForNgreater than 4 to 8 nodes, most real-world networks employ partial mesh connectivity.
46
ATM & MPLS Theory & Application: Foundations of Multi-Service NetworkingFigure 4-7. Partial mesh network