Data Communications Networking Devices - 4th Ed.pdf

2.4 PACKET SWITCHING NETWORKS _______________________________________________ 195Table 2.2Comparing X.25 and frame relayPerformance/operational feature X.25 Frame relayPerforms packet sequencing Yes noPerforms error checking yes noPerforms ¯ow control yes no, drops frameswhen congestionoccursNetwork access 300 bps±64 kbps 56 kbps±2.048Mbps, with T3access beingintroducedSwitch delay 10±40 ms 2±6 msOne-way delay 200±500 ms 40±150 msUtilizationThis also means that, by providing users with the ability to burst to hightransmission rates, the physical access to a frame relay network can be much higherthan access to an X.25 network.Since a frame relay network does not have to perform packet sequencing norerror checking, the ¯ow of data through a network is considerably faster thanthrough an X.25 network. Similarly, a frame relay switch will have a lower delay orlatency, and the ability to process more packets or frames per unit time. Table 2.2provides a general comparison between X.25 and frame relay network performanceand operation.In examining the entries in Table 2.2, the difference in one-way delay betweenframe relay and X.25 networks resulted in a new application for frame relay thatwould be impossible to perform effectively on older X.25 networks. This applicationis the transmission of voice, which is very susceptible to network delays.Although we will discuss voice over frame relay later in this section, we will defer adetailed examination of how FRADs enable voice to be transported until wediscuss the operation of this networking device later in this book.Frame relay represents a high speed internetworking technology well suited forconnecting geographically separated LANs and mainframes on an any-to-anyconnectivity basis via the mesh structure of public frame relay providers. Figure2.16 illustrates an example of how a public or private frame relay network could beused to interconnect three LANs and two mainframe computers. Note that a framerelay compliant router and front end processor requires only one connection to thenetwork to obtain the ability to communicate with multiple destinations. Toaccomplish this, private virtual circuits PVCs) must be established for eachlocation that requires access to another location. For example, if the front-endprocessor connected to the S/390 mainframe requires the ability to communicatewith each LAN, then three PVCs must be established.