14 Introduction and Layered Network Architecture Chap.1 the maximum delay requirement is stringent,and the need for reliability is rather low.A network that can handle all these applications together will probably not have too much difficulty with the other applications of interest. 1.2.3 Circuit Switching and Store-and-Forward Switching There are two general approaches,known as circuit switching and store-and-forward switching,that can be used within a subnet to transmit the traffic for the various sessions. A brief overview will be given of the circuit switching approach.followed by the reason why this approach leads to inefficient utilization of the communication channels for many types of sessions.Next,an overview of the store-and-forward approach will be given, showing how it overcomes the above inefficiency. For the circuit switching approach,when a session s is initiated,it is allocated a given transmission rate rs in bits per second (this could be different in the two direc- tions of a two-way session,but we focus on a one-way session here).A path is then created from the transmitting site through the subnet and to the destination site.Each communication link on this path then allocates a portion rs of its total transmission ca- pacity in the given direction for that session.This allocation of transmission rates to different sessions on a communication link is usually done by time-division multiplexing (TDM)or frequency-division multiplexing (FDM),but the details of that are explained in Section 2.1.What is important is that the sum of the rates for all the sessions using a link cannot exceed the total capacity of the link.Thus,if a communication link is fully allocated to existing sessions,a new session cannot use that link.If no path can be found using links with at least rs bits/sec of unused rate,the new session must be rejected (i.e.,given a busy signal).The other important point is that once the session has been successfully initiated,it has a guaranteed transmission rate r through the network. The nodes then simply take the incoming bit stream for a given session off the incoming link and switch it to the allocated portion of the outgoing link.This type of switching is quite similar to the well-developed technology for switching in the telephone network. In the telephone network,however,each session is allocated the same transmission rate, whereas in a data network,the required transmission rates are different and vary over a wide range. Circuit switching is rarely used for data networks.In the past,the reason for this has had nothing to do with the potential complexity of the switching,but rather,as we now explain,has been because of very inefficient use of the links.Typical data sessions tend to have short bursts of high activity followed by lengthy inactive periods;circuit switching wastes the allocated rate during these inactive periods.For a more quantitative view,let A be the message arrival rate for a given session s.More precisely,1/A is the expected interarrival time between messages of s.Let X be the expected transmission time of a message over a given link in the path;that is,if L is the expected length (in bits)of messages from s,and rs is the bit rate allocated to s,thenX=L/r.Figure 1.5 illustrates these arrivals and transmission times. Note from the figure that the fraction of time in which session s's portion of the link is actually transmitting messages is rather small:that portion of the link is otherwise14 Introduction and Layered Network Architecture Chap. 1 the maximum delay requirement is stringent, and the need for reliability is rather low. A network that can handle all these applications together will probably not have too much difficulty with the other applications of interest. 1.2.3 Circuit Switching and Store-and-Forward Switching There are two general approaches, known as circuit switching and store-and-forward switching, that can be used within a subnet to transmit the traffic for the various sessions. A brief overview will be given of the circuit switching approach, followed by the reason why this approach leads to inefficient utilization of the communication channels for many types of sessions. Next, an overview of the store-and-forward approach will be given, showing how it overcomes the above inefficiency. For the circuit switching approach, when a session s is initiated, it is allocated a given transmission rate T s in bits per second (this could be different in the two direc￾tions of a two-way session, but we focus on a one-way session here). A path is then created from the transmitting site through the subnet and to the destination site. Each communication link on this path then allocates a portion Ts of its total transmission ca￾pacity in the given direction for that session. This allocation of transmission rates to different sessions on a communication link is usually done by time-division multiplexing (TOM) or frequency-division multiplexing (FOM), but the details of that are explained in Section 2.1. What is important is that the sum of the rates for all the sessions using a link cannot exceed the total capacity of the link. Thus, if a communication link is fully allocated to existing sessions, a new session cannot use that link. If no path can be found using links with at least T s bits/sec of unused rate, the new session must be rejected (i.e., given a busy signal). The other important point is that once the session has been successfully initiated, it has a guaranteed transmission rate T s through the network. The nodes then simply take the incoming bit stream for a given session off the incoming link and switch it to the allocated portion of the outgoing link. This type of switching is quite similar to the well-developed technology for switching in the telephone network. In the telephone network, however, each session is allocated the same transmission rate, whereas in a data network, the required transmission rates are different and vary over a wide range. Circuit switching is rarely used for data networks. In the past, the reason for this has had nothing to do with the potential complexity of the switching, but rather, as we now explain, has been because of very inefficient use of the links. Typical data sessions tend to have short bursts of high activity followed by lengthy inactive periods; circuit switching wastes the allocated rate during these inactive periods. For a more quantitative view, let'\ be the message arrival rate for a given session s. More precisely, 1/,\ is the expected interarrival time between messages of s. Let X be the expected transmission time of a message over a given link in the path; that is, if I is the expected length (in bits) of messages from s, and Ts is the bit rate allocated to s, then X = I/Ts . Figure 1.5 illustrates these arrivals and transmission times. Note from the figure that the fraction of time in which session s's portion of the link is actually transmitting messages is rather small; that portion of the link is otherwise