|
First released in 1974, MVS™ (Multiple Virtual Storage) and its descendants remain the most commonly used operating systems on IBM mainframes. In this paper, Auslander, Larkin, and Scherr trace the evolution of MVS and describe the underlying technological advances from the first IBM operating systems in the mid-50s through the 1970s.
The environment in which MVS evolved was ruled by a number of constraints. First, because of the huge investment made by customers in application programs, the operating system had to ensure compatibility with previous systems by preserving the operational characteristics of their interfaces. The operating system had to support a multitude of hardware configurations that included a series of processing units with different speeds and multiprocessing capabilities, a range of main storage sizes, a varying number of I/O channels, and various device types. Due to the large number of installed mainframes, which means that the deployed systems experienced extreme workloads and a wide range of applications, any hardware or software defect stood an increased chance of causing a failure. Because customers demanded stable, reliable systems, this stress environment represented a tall challenge for system designers. They rose to that challenge, and to this day, the descendants of MVS running on IBM mainframes remain technology leaders in security, reliability, and performance.
The authors describe the historical trends they observed. Whereas at the dawn of the computer era manual operation was the norm, automation gradually took over, and as a result, the stand-alone batch system of the 1950s was replaced by the geographically dispersed but centrally managed multisystem complex in the late 1970s. Facilities commonly used by various applications were moved into the operating system. The level of the interface for the application programmer was raised to mask the details of I/O-device and hardware operation. Not only was there an order-of-magnitude improvement in reducing the number of errors in deployed operating systems, but the impact of these errors was reduced by new recovery strategies in the software itself. Most significantly, the authors correctly predicted that the steady decrease in the cost of hardware together with the expected increase in the cost of human labor would continue to shape the design of computing systems for the foreseeable future.
|
