An integrated approach to automated computer maintenance

This paper describes a method for applying automatic fault locating techniques to a computer system. Although the techniques described here are generally applicable, they have been specifically applied to IBM System/360 Model 50. Hardcore considerations are stressed and levels of hardcore are defined. Implementation objectives include both minimization of hardcore and minimization of additional circuitry. The design philosophy is to alter existing hardware where possible, keeping hardcore restrictions in mind. Hardware is added only where function does not exist or where hardcore restrictions are violated. A bootstrap approach is used whereby subsystems are tested before being used for checkout of other subsystems. A special program load is used to bypass CPU and channel circuitry to enter tests into storage. With tests in storage, a mode of operation is defined to permit use of main storage as both a data and control storage. In this mode, tests are performed on the read only storage controls. The read only storage, once checked, is used to control test execution on the remainder of the CPU. The entire CPU is then used as a test vehicle for channel checkout purposes. Finally, the same techniques and circuitry are shown to be applicable to other maintenance functions, including logging of errors and checking of certain storage subsystems.