THINK Home Digital Archive Search Feedback



A Technical History of National Physical Laboratories (NPL) Network Architecture - A Technical Tour



Links in the network are over connected. The use of redundant links along with alternative routing ensures that the network will maintain service event in the event of equipment failures. It must be noted that packets will be lost when links becomes non-operational. Therefore, interface computers provide methods of ensuring the quality of service provided by the network. The links in the network does not implement a standard bit rate. Each individual link has its own standard bit rate for data transfer, but the rates varies from link to link. Although the rates are non-standard, Davies' initial calculations estimate that the network will average a minimum of 1.5Mbps.

Mark I Networks

The experimental Mark I network was originally designed to have three packet switching nodes. It was to act as a switching computer which would make virtual connection between any two devices. But funding constraints reduced the number to one. With only one node, the NPL team would not get the opportunity to tackle issues like congestion and routing. The node computer used was the Honeywell 516. The Mark I had about sixty lines that provided access to a DEC PDP-8 computer and two mainframes. Through the network, NPL researchers could have remote access to computers for writing and running programs, querying a database, sharing files, special services such as a "desk calculator", and for simple communication between users. The network provided a throughput of about 150 packets per second (much less than the couple thousand per second forecasted by Davies early on). In its first form the Mark I lacked a packet interface; it was just a byte interface which would produce a communication link between two terminals, but it wouldn't enable one computer to talk to several terminals by interleaved packets. The Mark I was only a single character switch rather than a packet switch. In addition, the Mark I network operating system proved to be unwieldy and slow, largely due to the aforementioned hardware implementation. (Campbell-Kelly)

During the 1970's, additional equipment was added to the network to attract users. In 1971, the network was fully functional. (Campbell-Kelly)

The Mark I


Mark II Network

The upgraded Mark II system, introduced in 1973, used most of the same hardware as the Mark I, but software and protocol improvements made it two to three times faster. In order to eliminate the problem of "packetization of data that had made certain kinds of terminal interactions impossible" by the Mark I network, the Mark II used an "intelligent terminal processor" to act as a packet assembler/disassembler. In addition, the Mark II provided for both the character switched and the packet switched interface. The new packet switched interface allowed for user machines, i.e. hosts, to attach to the network (as planned by Davies in his 1955 proposal). The Mark II also differed from it's predecessor because it begain to take a layered approach versus a structured scheme. With all of the improvements in place, the upgrade network had maximum packet rate of 250-500 packets per second.

The Mark II remained in service at NPL until 1986- quite an impressive term of service for an experimental system. (Campbell-Kelly)





Written by the THINK Protocols team, CS Dept, UT Austin
Please direct comments to Chris Edmondson-Yurkanan.

This document was last modified on Tuesday, 11-Jun-2002 10:18:01 CDT.