Principles of Broadband Switching and Networking

3

FUNDAMENTAL PRINCIPLES OF PACKET SWITCH DESIGN

The next few chapters focus primarily on fixed-length packet (or cell) switching. To switch variable-length packets using a fixed-length switch, one could first fragment the variable-length packets at the inputs into small fixed-length packets, switch the packets, and then reassemble the packets back into the original variable-length packets at the outputs. Generally, it is possible for the packet boundaries of different inputs to be unaligned (see Fig. 3.1). Because of the fixed packet length, we can deliberately delay the early arriving packets (e.g., store them into a buffer) in order to align the boundaries. Thus, as far as the operation of the switch is concerned, we may assume that time is divided into fixed slots of one-packet duration. In the beginning of each time slot, new packets arrive at all the inputs simultaneously.

One of the exercises in the preceding chapter examines time–space–time circuit switching from the viewpoint of contention resolution. Specifically, the information on different inputs may be destined for the same output of the space-division switch at any given time slot (see Fig. 2.23). The input TSI rearranges the time slots occupied by the information so that output conflicts do not occur.

In circuit switching, the time slots occupying the same relative position in successive frames of an input are assigned to the same circuit, and they contain information destined for the same output. In other words, ...