Software Related Failures on the IBM 3081: A Relationship with System Utilization

摘要

This paper presents an analysis of software-related system failures on the IBM 3081 at SLAC. We find three broad categories of failures: error handling, control or logic problems and hardware-related. A statistical analysis shows (not unexpectedly) a decreasing failure rate with time, This is especially true in the early part of the study. Not withstanding the decreasing failure with time, we find that the occurrence of failures is strongly correlated with the type and level of workload prior to the occurrence of a failure. For example, it is shown that the risk of a software-related failure increases in a non-linear fashion with the amount of interactive processing, as measured by paging rate and system overhead. The paper employs a statistical model to describe the load dependency and offers explanations for the observed phenomenon. (Author)