|
|
|
Two different switching techniques are used inside the telephone system:
When a user place a telephone call, the switching equipment within the telephone system seeks out a physical "copper" (including fiber and radio) path from the caller telephone to the callee telephone. This technique is called circuit switching (Fig. 2-34(a)).
Fig. 2-34. (a) Circuit switching. (b) Packet switching.
An important property of circuit switching is the need to set up an end-to-end path before any data can be sent. It takes some set-up time during which there is no data transmission in progress. Long set-up times are for many computer applications undesirable.
As a consequence of the path between the calling parties, once the set-up has been completed, the only delay for data is the propagation time for the signal and there is no danger of congestion.
An alternative switching strategy is message switching (Fig 2-35(b)). In this case, no physical copper path is established in advance. Instead, the store-and-forward technique for the entire messages is applied. It was first used for telegrams.
Fig. 2-35. Timing of events in (a) circuit switching,
(b) message switching, (c) packet switching.
With message switching, there is no limit on block size, which means that routers must have disks to buffer long blocks. It also means that a single block may tie up a router-router line for minutes, rendering message switching useless for interactive traffic.
To get around these problems, packet switching was invented. Packet-switching networks place a tight upper limit on block size. So no user can monopolize any transmission line very long and therefore these networks are well suited to handle interactive traffic. A further advantage of packet switching over message switching is (Fig. 2-35(c)) that the first packet of a multipacket message can be forwarded before the second has fully arrived, reducing delay and improving throughput. For these reasons, computer networks are usually packet switched, occasionally circuit switched, but never message switched.
The differences between circuit switching and packet switching are summarized in Fig. 2-36.
Fig. 2-36. A comparison of circuit-switched and packet-switched networks.
As an example of circuit-switched telephone system we will briefly describe the AT&T system. The system of other companies or countries have the same general principles.
Fig. 2-37. The AT&T telephone hierarchy. The dashed lines are direct trunks.
The basic rules of operation of the system are (Fig. 2-37):
The crossbar switch (Fig.2-38) is the simplest kind of switch. In case of n input lines and n output lines it has n2 crosspoints, where input and output lines can be connected by semiconductor switches.
Fig. 2-38. A crossbar switch with no connections. (b) A crossbar
switch with three connections set up:
0 with 4, 1 with 7,
and 2 with 6.
The problem with a
crossbar switch is that the number of crossbars grows as the square of the
number of lines into the switch. If we assume that all lines are full duplex
and that there are no self connections, only the crosspoint above the diagonal
are needed. Still, n(n - 1)/2 crosspoints are needed. For n=1000, we need
499500 crosspoints. It possible to build a VLSI chip with this number of
transistor switches, but not with 1000 pins on the chip. Thus a single
crossbar switch is only useful for relatively small end offices.
2.4.15. Space Division Switches
By splitting the crossbar switch into smaller ones and interconnecting them, it is possible to build multistage switches with fewer crosspoints. These are called space division switches. Two configurations are illustrated in Fig. 2-39.
Fig. 2-39. Two space division switches with different parameters.
The number of crosspoints needed for a three-stage switch is 2kN + k(N/n)2. For N =1000, n = 50, and k = 10, we need only 24000 crosspoints instead of the 499500 required by a single-stage crossbar.
However, this type of switch can make much less connections at the same
time comparing with the single-stage crossbar (8 in case (a), 12 in case (b)).
2.4.16. Time Division Switches
With time division switch (Fig. 2-40), the n input lines are scanned in sequence to build up an input frame with n slots. Each slot has k bits. For T1 switches, the slots are 8 bits, with 8000 frames processed per second.
Fig. 2-40. A time division switch.
The heart of the time division switch is the time slot interchanger, which accepts input frames and produces output frames, in which the time slots have been reordered according to mapping table in the memory of the switch. Finally, the output frame is demultiplexed with output slot 0 going to line 0, and so on. In essence, the switch moves data from input lines to output lines according to the mapping table even though there are no physical connections between these lines.
The problem that limits the number of input lines to a time division switch is the time necessary to transform an input frame into the corresponding output frame. It is necessary to store n slots in the buffer RAM and then to read them out again within one frame period of 125 (sec. With memory access time T, we need a time interval 2nT, so with T = 100 nsec we can support at most n = 125/2T = 625 lines.