510 Part 5 The PMS level Section 4÷ Network computers and computer networks

(10 char/s) and medium (1,200 ~ 4,800 b/s) speed, as shown in Fig. 6a. In this regard the network is simply a message switch for the three terminal types. It employs Cs for the switching elements and is fundamentally a store-and-forward system. Had it not been for security, reliability, response time, and other considerations, it would have been possible to construct an equivalent system using standard lease wire switches (or telephone exchanges). In Fig.

6b a tree is used to present the relationship of constituent members of ComLogNet. From it we see that at the first level ComLogNet has just a switch, links, and terminals (as shown in Fig. 6a). The network's switch employs five specialized N('Automatic Electronic Switching Centers/SC)'s which communicate among each other (Fig. 6c). Terminals connect to the individual N('SC)'s and messages are routed between two T's, either by a store-and-forward process within NQSC) or among two N('SC)'s.

The individual N('SC)'s are located at five specific locations and consist of fixed computer configurations of five to seven C's. The structure of N('SC) (Fig. 6d) is formed basically by a duplex C structure which handles most processing. Attached to the two C('Communications Data Processor/CDP) are two to four C('Accumulation and Distribution Unit/ADU) which handle communication-link processing. A C('Tape Search Unit) is used off line to process data from Ms(magnetic tape). The structures of C('CDP), C('Tape Search Unit), and C('ADU) are defined within Fig. 6d.

An experimental computer network (Fig. 7a) is operational and connects 19 computer facilities associated with the contractors of the Information Technology Branch of the Advanced Research Projects Agency (ARPA). These contractors, all of whom are engaged in advanced research in computer science and technology, form a community in which to attempt a general-purpose network. Since several of the nodes in this network (e.g., M.I.T.; see Fig. 4) will themselves be constructing networks at their own sites, the system has faced a good many of the design problems associated with such a network. Unlike many of the other networks discussed in this chapter, the ARPA network consists of sites that are physically remote, that are each developing as total systems under independent management, and that have no agreed-upon functional specialization vis-a-vis each other. Furthermore, the uses that each node will make of other nodes will be the fairly general ones cited at the beginning of this chapter, as generated by a general scientific community. Since many of the institutions that will be tied in are major academic institutions, diversity will be guaranteed. The motivation behind the experiment is to reveal and begin to solve the technical problems of such general networks, while also discovering which of the several advantages of using networks listed earlier (or others unmentioned) emerge as important.