Today's lesson compares link state routing with distance vector routing, and further, discusses ATM networks. Many routers in the Internet support both distance vector routing and the link state routing protocols. Within a domain, nodes must cooperate and exchange information to obtain the values of the metrics. In the distance vector routing, neighboring routers exchanging routing tables. Is used to see whether there are new, better paths to the neighbor that provide the information. The problem is it adapts to change in the network topology slowly. On the other side, the reaction to failure in link state routing is very quickly. If a link fails, the router sets the link distance to infinity and floods the network with an update packet. All routers immediately update their link database and recalculate their shortest paths. Link state routing has many other benefits. It can support for multiple metrics at once. Link state routing also supports for multiple paths to the destination, and even more flexible for source routing. But the shortcoming of link state routing is the old duplicate messages due to flooding. We have discussed several measures to deal with it, for example, adding a sequence number to each updated message, or adding a time to live to each message. In the datagram network, each router is responsible for determining the next hop along the shortest path. If every router performs the same process, it is hop-by-hop routing. Source routing is another routing approach, whereby the path to the destination is determined by the source router. Source routing allows the host to control to the path that its information traverse in the network. The figure shows an example. The source host initially includes the entire path, 1, 3, 6, B, in the packet to be designated to host B. Each node examines the packet header, strips off the address identifying the node, and forwards the packet to the next node. Asynchronous transfer mode (ATM) is a method for multiplexing and switching that provides rich quality of service support. It applies fixed-length packets, called "cells". Each cell is 53 bytes longs, including 5 bytes header. ATM is connection-oriented, and therefore conceived as end-to-end. It supports a broader range of services, so real time voice and video, circuit emulation for digital transportation, and data traffic with bandwidth guarantees. ATM combines several desirable features of packet switching and time-division multiplexing, TDM, second switching. The table compares full features of TDM and of packet multiplexing. The first comparison involves the capability to support services that generate information at a variable bit rate. The second comparison involves the delay incurred in traversing the network. The third comparison involves the capability to support burst traffic, and the fourth comparison involves the processing time. It operates as the figure shows. The information flows generated by various users are converted into cells and sent to an ATM multiplexer. The multiplexer arranges the cells into one or more queues, and implements some scheduling strategy that determines the order in which the cells are transmitted. The purpose of the scheduling strategy is to provide for the different quality of service required by the different flows. ATM doesn't reserve transmission slots for specific information flows, and so it has the efficiency of packet multiplexing. The reason for the term asynchronous is that its transmission of cells is not synchronous to any frame structure. ATM networks are connection-oriented and require a connection set up prior to the transfer of cells. The connection setup procedure is similar to that described for virtual circuit packet switching networks. The connection is called a virtual channel connection that is established by a chain of locally-defined tags. Also, ATM has a strong similarity to virtual circuit packet switching. One major difference is ATM use of short, fixed-length cells. This approach simplifies implementation of switch, and it makes high speed possible. ATM was initially considered as more scalable than packet switching. It was envisioned as a high-speed, multi-service network technology for the core packet network. However, the overhead in the ATM cell header was found to be a significant penalty in link efficiency. The need to implement a segmentation and reassembly for ATM also proved challenging. Multiple protocol label switching, MPLS, was introduced as an alternative network technology for the core packet network. It adopts the label switching paradigm, as ATM, that allows variable-length packets.
