First-Hand:Event in Telecom Switching Development
Events in Telecommunications Switching Development: No. 2 ESS and ITT System 12
By Philip C. Richards
Two major events in my career:
1. Between 1965 and 1973, with the early development of Electronic Switching Systems (ESS) by Bell Laboratories, the primary electronics were the computers that controlled the system and the control networks. The actual switching and network interfaces were electronic mechanical using reed switches for the network matrix and relays for network control interfaces to trunk circuits and lines. Previous all electromechanical systems used a multiplicity of relay-based control systems called markers to route calls. The ESS design used two computers working in lockstep, running the same programs that received identical inputs. A special circuit matched the outputs of the two computers, and any differences caused the system to momentarily suspend operation until the test programs could determine which computer had failed.
The primary assumption was that the cause of the error was hardware. However, experience showed that frequently the problems were due to software errors. I was a key designer of the No. 2 ESS, which benefited from the early experience with the No. 1 ESS in having software failures. Our design approach started with the assumption that the larger source of errors would be from software. As a result, we developed many tools to detect potential software errors before they brought down the system. Indeed, we had in place several teams of experts ready to respond to any hints of problems. To our great relief, these teams weren't called upon, and the very first systems operated without stopping even once starting in November, 1969 for the longer than the a year in serving customers. This was accomplished by several techniques:
a. Several special programs were included in the system specifically to allow fixing software by changing the program in one computer while the other computers handled the calls. Then running the new computer program in one computer along side the old program in the other computer to determine if there were any flaws. Then the new program took over, and the old program was updated in the second computer before it was restored to the standard parallel operation.
b. During normal operation, when the computers were not busy handling calls, the software was constantly auditing the memory structure to eliminate any inconsistencies. Any inconsistencies found were highlighted as potential software errors before they could affect the whole system operation.
c. Very thorough testing was done by having two independent systems working together with one generating telephone calls to be processed by the other system. Prior to this time, devices called load boxes were used for this purpose but didn't have the capacity to fully test the load carrying capability of the system. Using these back-to-back systems, the system under test could be overloaded to see the actual results - both to know what the real capacity was and also how the system reacted to heavy load.
The focus on absolute protection of operation was a hallmark of these computer controlled systems which demonstrated that the objective of no more than two hours of downtime in forty years was quite realistic even though most of them were replaced by later generation systems before the forty year was up. In light of today's computer systems where outages of hours several times a year are tolerated, the ESS systems clearly demonstrated that continuous operation can be obtained and interruptions should not be tolerated.
1970 - Bell System Technical Journal - No. 2 ESS Call Processing Software
2. In April, 1975, I was tasked to plot the future of digital switching development for the worldwide supply of telephone switching systems from the many companies that were part of the ITT conglomerate. While most of these companies were located in Europe, other were is North America and Asia. Each company supplied their domestic system demand as well as providing exports to a number of countries. The requirements for a corporate-wide system solution were many and varied, meaning that beyond a core solution, adaptations to local requirements had to be readily developed.
For the core structure, it was apparent that just adapting existing architectures to the needs of digital switching would create a massive development effort, with each country's solution based on the many and varied prior generation systems. Instead, I conceived of a revolutionary approach that benefited from the rapid advance in technology. Indeed, the concept would not become practical until the end of the four years that it would take to develop the system.
At the time, several other digital switching systems were in various stages of development, all with architecture that were based on the centrally control structure of the earlier Electronic Switching Systems, while using a digital switching matrix that used two time division switching matrices interconnected by a space division switching matrix. The result was an interconnecting network that either had limited expandability or was too cumbersome because of the allowance for growth.
I was very well aware of the very first telephone switching system call Step-by-Step which was invented in the late 1800's, and continued to serve many communities even in this latter quarter of the 1900's. Because the switching was directly controlled by the dialing of the user, no major control system was required. However, this resulted in very limited flexibility to adapt to new requirements. As a consequnce, more complicated systems came into use that relied on centralized control systems, first based on relays, and later based on computers. However, these systems were constrained by their structure with limitations on size, both physical and economic -- either limited to smaller locations or much larger locations.
In the spring of 1975, I conceived of a solution that was loosely based on the network structure of Step-by-Step switching, but taking advantage of the digitalization of the phone calls, with each stage in the network using a multiplicity of small time and space digital switches, interconnected to cascading stages of additional digital switches, all controlled from the ends of the network by initial control signals that established a path through the several switching stages. This same network would also allow control signals to be sent to establish the routing of the call, resulting in a structure that required no centralized control system allowing for Fully Distributed Control. This control structure also took advantage of the evolution of technology by using hundreds of microprocessors to handle the control tasks. The structure also allowed the size of the system to work economically for very small networks serving a thousand lines to very large networks serving more that one hundred thousand lines.
At the time this was conceived, the combined space and time division switching elements would have been impractical. However, I projected that four years later, each of these elements could be implemented on a single integrated circuit chip. This projection indeed came true and the network elements were implemented as single chips. After completion of the development, this became known as System 12 from the various ITT supply houses, and over the next fifteen year, became the largest selling switching system in the world.
1979 Third World Telecommunications Forum (Geneva) - ITT 1240 Digital Switching System