Design of Serviceability Features for the IBM System/360

Abstract

This paper discusses the design of features that are intended to provide the IBM System/360 with u significant improvement in serviceability over that of previous systems. It was decided from the beginning to develop the System/360 as an integrated package of hardware, operational programs, and maintenance procedures. The major problems to be solved in gaining this improvement and integration were (a) reducing the maximum duration of service calls; (b) reducing the median duration and mean duration of service calls; and (c) matching a single package of maintenance programs and procedures to a large variety of operational monitor programs and machine models. These problems have been attacked by supplementing standard servicing facilities (both hardware and program) with (a) the ability to record automatically the complete, detailed, system environment at the instant of error discovery; (b) the ability to initialize the CPU to any arbitrarily specified state (either legal or illegal), to advance from this state by a specified number of machine cycles, and to compare the new state with a precomputed result state, much of this using circuits that are independent of those required for program sequencing; (c) a system of programs that can be integrated with the System/360 Design Automation to produce automatically the inputs, results, and location analyses that are required to exploit the capabilities described in (b); (d) a family of diagnostic monitor programs that attack directly the problem of matching maintenance procedures to machine models and operational monitor programs; and (e) a facility to retry failing CPU operations at the instruction level in the larger models, in addition to the usual retry at the program-segment level.