Some Problems in Automating Fault Control in Electronic Digital Computers

V. M. Razin, Yu. N. Efimov, M. N. Strogonov

(Presented by the scientific seminar of the Faculty of Automation and Computer Engineering)

Modern radio-electronic devices for obtaining and processing information are complex systems that perform highly important tasks. Forced downtime associated with carrying out current repairs to eliminate sudden failures leads to a substantial reduction in the operating efficiency of complex and expensive equipment. Average losses in systems for processing discrete information of central importance with a continuous process may in many cases be regarded as proportional to the downtime [1]:

\[ W(t_p)=A_0 t_p, \]

where \(A_0\) is the specific loss (loss per unit time) during system downtime, equal to the average revenue that the system would have produced in the case of proper operation;

\(t_p\) is the downtime caused by the elimination of catastrophic failures.

In modern electronic digital computers with a speed on the order of 50 thousand operations per second, the value \(A_0 \cong 20\) thousand rubles per hour, if it is assumed that the same computational work is performed manually with the aid of keyboard calculating machines. It is therefore clear that even short unscheduled downtimes lead to large economic losses. A significant share of the time spent on repairs caused by sudden failures is taken up by the search for the failed unit or element.

When a modular design is used, the time spent on the actual repair is relatively small, since it reduces to replacing the detected faulty module with a serviceable one. The problem of locating a failed module (subunit) in modern complex digital computers is difficult and still far from completely solved.

Built-in automatic self-checking systems, used in some complex radio-electronic information-transmission devices [2], often prove to be of little suitability for digital computers, which usually have a complex, branching logical structure. An attempt to construct automatic control of all circuits of a computing machine leads to a significant

(more than doubling) the complexity and composition of the machine’s equipment.

Test-control methods, widely used in the maintenance practice of electronic digital computers, for the most part make it possible to establish the type of operation being performed incorrectly, but do not make it possible to determine the location of the fault. Diagnostic tests make it possible to solve this problem only partially and are far from universal; for some types of damage they are not applicable [3].

Apparently, more satisfactory solutions can be obtained by using combined methods of circuit and test control with the appropriate participation of the operator [3, 4].

In light of these considerations, the work [5] is of interest; in it, methods are specified to some extent for controlling one digital electronic computer with the aid of another serviceable machine. The fundamental possibility of such control was noted in the above-mentioned work [3].

The essence of the proposal in [4] amounts to the fact that, with the aid of the serviceable machine, not all elements of the faulty machine are checked, but only certain ones, called basic elements.

Further checking of the faulty machine is carried out on the basis of using the already checked and repaired basic elements in a strictly defined sequence, with the use of self-checking principles.

This promising method has shortcomings, consisting in the fact that another machine is not always available that can be used to check the faulty machine and, if such a machine is available, such a check is not always expedient.

When faults are detected in the basic elements, repairs must be carried out, after which these basic elements must be subjected to a repeated check confirming their serviceability. This procedure, when there is a large number of faults, may entail large expenditures of machine time of the checking machine.

In order to overcome the indicated difficulties, it appears expedient to construct a specialized logical computing machine of discrete action in the form of an attachment intended only for checking the minimum number of basic elements. Such an attachment must have relatively small dimensions, in view of the small number of elements being checked, must be sufficiently reliable, and must have self-checking circuits.

In this case it is assumed that the machine being checked must have a built-in automatic-control system covering some of the elements whose operation can be checked without substantially complicating the machine itself. What is meant here are power supplies, generators of periodic clock pulses, power amplifiers, etc. The system of diagnostic tests in this case must include checking only those elements that are not covered by the automatic-control system.

Such a combined approach to solving the problem of finding faults in electronic digital computers will not lead to a significant increase in the composition of the equipment and will allow the operator, with minimal time expenditure, to localize effectively, if not all, then a considerable part of the faults arising in the course of operation.

References

1. G. A. Shastova. A criterion of average losses for evaluating the reliability of a control system. Avtomatika i telemekhanika, No. 6, 1962.

2. A system of automatic monitoring and troubleshooting in radio-electronic equipment (survey). Zarubezhnaya radioelektronika, No. 5, 1961.

3. E. I. Klyamko. Schematic and test monitoring of automatic digital computers. Sovetskoe Radio Publishing House, 1963.

4. D. S. Zhilkin, Ya. A. Khetagurov. Some problems of hardware–logical monitoring of units of an electronic computer. Vychislitel’naya tekhnika, issue 4, Gosatomizdat, 1962.

5. A. K. Olefir. On the search for faulty elements in a system of two computers. Computing Systems. Collection of works of the Institute of Mathematics, Siberian Branch of the USSR Academy of Sciences, issue 6, 1963.