Program-counter-based prediction in operating systems

Instructions uniquely identified by the program counters provide the context of program execution and instruction-based prediction techniques have been widely used at the architectural level. Operating system researches, on the other hand; have not explored the benefits of instruction-based prediction for resource management. This research explores the potential benefits provided by instruction-based prediction in operating systems. In particular, we investigate the potential of using instruction-based prediction techniques for managing I/O devices in operating systems. The thesis will first propose instruction-based classification technique for buffer cache management. Our technique classifies and predicts the I/O access patterns in the applications, allowing the buffer cache manager to apply the best cache replacement scheme for each predicted access pattern. The second part of the thesis proposes an instruction-based technique for power management that dynamically learns the access patterns and associated idle times of applications to predict when an I/O device can be shut down to save energy. The technique uses path-based correlation to observe a particular sequence of I/O triggering instructions leading to each idle period, and predicts future occurrences of that idle period