previous | contents | next

Chapter 28½ Fault-Tolerant Design of Local ESS Processors 469


gram interrupt, which reads the word from the standby store in an attempt to bypass the error.

As discussed previously, the No. 2 Processor (first generation) is used in the No, 2 ESS for medium-size offices, It covers approximately 4000 to 12 000 lines, with a call handling capability of 19 000 busy-hour calls. (The number of calls is related to the calling rate of lines during the busy hour.) The microprogram technique used in the No. 3A Processor design allows the No. 2 Processor's instruction set to be emulated. This enables programs written in the No. 2 assembly language to be directly portable to the No, 3A Processor. The ability to preserve the call processing programs permits the No. 2 ESS to be updated with the No. 3A Processor without having to undergo a complete, new program development.

The combination of the No. 3A Processor and the peripheral equipment of the No. 2 ESS is designated as the No. 2B ESS. It is capable of handling 38 000 busy-hour calls, twice the capability of the No. 2 ESS [Mandigo, 1976], The No. 2B ESS can be expanded to cover about 20 000 lines. Furthermore, when an existing No. 2 ESS system in the field exceeds its real-time capacity, the No. 2 Processor can be taken out and replaced with the No. 3A Processor. The retrofit operation has been carried out successfully in working offices without disturbing telephone service.


Summary

In order to achieve the reliability requirements, all ESS subsystem units are duplicated. When a hardware failure occurs in any of the subunits, the processor is reconfigured into a working system around the defective unit. The partitioning of subsystem units into switching blocks varies with the size of the ESS processors. For the medium- or small-size processors such as the No. 2 or the No. 3, the central control, the main memory, the bulk memory, and the store bus are grouped as a single switchable entity, A failure in one of the subunits is considered a failure in the switchable block. Since the number of components within a switchable block is sufficiently small, this type of single-unit duplex configuration meets the reliability requirement. For larger processors such as the No. 1 or the No. 1A, the central control, the program store, the call store, the store buses, and the bulk file store are treated individually as switchable blocks. This multiunit duplex configuration allows a considerable number of combinations in which a working system can be assembled. The system is down only when two simultaneous failures occur, one in the subunit and the other in the duplicated subunit. A greater fault tolerance is possible with this configuration. This type of configuration is necessary for the large processor because each subunit contains a larger number of components.

The first generation of ESS processors, which includes the No. 1 and the No. 2, have provided commercial service since 1965 and 1969, respectively. The No. 1 ESS serves large telephone offices (metropolitan); the No. 2 is used in medium-size offices (suburban). Their reliability requirements are the same. Both processors depend on integrated maintenance software, with the hardware that must (1) quickly detect a system failure condition, (2) isolate and configure a working system around the faulty subunit, (3) diagnose the faulty unit, and (4) assist the maintenance personnel in repairing the unit. The primary detection technique is the synchronous and match mode of operation of both central controls. Matching is done more extensively in the No. 1 than in the No. 2 since cost is one of major considerations in the design of the No. 2 Processor. In addition to matching, coding techniques, diagnostic access, and other check logic have been incorporated into the basic design of these processors to realize the reliability objectives.

The widespread acceptance of the No. 1 ESS and the No. 2 ESS has created the need for a second generation of ESS processors: the No. 1A and the No. 3A. They offer greater capability and are also more cost-effective. Both processors use the same integrated technology. The 1A Processor extends its performance range by a factor of four to eight times over the No. 1 Processor by using faster logic and faster memory. The 1A design takes advantage of the experience gained in the design and operation of the No. 1 ESS. The No. 1A Processor provides considerably more hardware for error detection and more extensive matching of a large number of internal nodes within the central control. The design of the No. 3A Processor had benefited by the experience gained from the No. 2 ESS. A major departure in the design of the 3A Processor from the design of other ESS processors is the nonsynchronous and the nonmatch mode of operation. The No. 3A Processor uses self-checking as primary means of error detection. Another departure is in the design of the No. 3A Processor's control section; it is microprogrammed. The No. 3A Processor's flexibility permits emulation of the No. 2 Processor quite easily.


References

Ault et al. [1977]; Ault, Gallaher, Greenwood, and Koehler [1964]; Becker et al. [1977]; Beuscher et al. [1969]; Bowman et al. [1977]; Browne, Quinn, Toy, and Yates [1969]; Budlong et al. [1977]; Cagle et al. [1964]; Chang, Smith, and Walford [1974]; Downing, Nowak, and Tuomenoksa [1964]; Fleckenstein [1974]; Genke, Harding, and Staehler [1964]; Goetz [1974]; Harr, Taylor, and Ulrich [1969]; Irland and Stagg [1974]; Keister, Ketchledge, and Lovell [1960]; Keister, Ketchledge, and Vaughan [1964]; Mandigo [1976]; Nowak [1976]; Seley and Vigilante [1964]; Smith [1972]; Spencer and Vigilante [1969]; Staehler [1977]; Staehler and Watters [1976]; Storey [1976]; Toy [1978]; Tsiang and Ulrich [1962].

previous | contents | next