How the Bell System Missed the Internet 3
“The Bell System ACS system Takes Shape…and Fails…”
This series discusses the largest private R&D development project in history as of the 1980s. It was a system called Advanced Communications System, or ACS. It was destined to become what the Internet is today.
Bell had examined existing technologies out there such as Tymnet and Compuserve and ARPANet. The basic problem with all of those was that they couldn’t scale up to the size needed by the massive Bell System.
At the time, the ARPANet system which eventually did become the internet was connected to only dozens of computers. It required very expensive IMP processors to actively switch packets. This system just didn’t scale up, it had too limited of throughput, it lacked maintenance features, and it was tightly controlled by BBN who preferred to feed off military budgets rather than work with private sector companies.
From the outset, the Bell System viewed their ACS system as a way to offload data traffic from the PSTN. It was a homogenous system that could talk to all sorts of systems out there. It could natively speak IBM 3270 bi-sync (the dominant terminal of the 1970s) as well as async, and even isochronous transmission. At the outset it would support HDLC and the emerging X.25. It was an ambitious design.
Above all else, ACS had to scale up to tens of millions of connections in a cost-effective way.
The decision was to use a highly distributed architecture. Line cards would service a limited number of ports (such as 8 or 16 serial ports or 4 synchronous ports.) The Line cards would interchange data across a time-space-time buss similar to what central offices of the day used, rather than to route traffic through one central processor. It was the routing of traffic through a main central processor like the ARPANet IMP which was seen as the main problem with cost and scalability.
This was not an entirely foreign project. Within the Bell System there was already a specialized Telex-like system that was cobbled together out of a #1ESS that provided telegram service for a mysterious customer. It must have been military because discussing this was always hushed. But the switch existed.
A design for ACS was settled upon that would use Western Electric processor and memory components. A prototype system was built and each of the components was individually tested.
The ACS project started with under 100 engineers and development progressed relatively smoothly. The architecture was finalized and sub-system design progressed. The line cards, and other parts of the system were developed and readied for production.
The staff continued to grow and it reached over 200 people within a year. Now, people involved with writing Bell System Practices were busy asking questions. Thousands of pages of documentation were written covering everything from how to install the system to how to manage and maintain it.
Eventually, It was time to integrate the system into one working piece and then to deploy it. This is the time that I got involved in the project full-time.
The problem was that ACS was starting to run late. It had missed several deadlines and managers were reluctant to make new commitments. By the time news was sent up 7 levels of management (the Bell System had exactly 10 levels of management) and then across departments, the news was so sanitized that it was untrustworthy. Billy Oliver couldn’t get at the truth because the people who knew were too distant and the people close didn’t know. My job was to cut through that.
Next…Fail…ACS transforms into a disaster…