Exploiting recurring usage patterns to enhance filesystem and memory subsystem performance

Abstract

In many cases, normal uses of a system form patterns that will repeat. The most common patterns can be collected into a prediction model which will essentially predict that usage patterns common in the past will occur again in the future. Systems can then use the prediction models to provide advance notice to their implementations about how they are likely to be used in the near future. This technique creates opportunities to enhance system implementation performance since implementations can be better prepared to handle upcoming usage. The key component of our system is the ability to intelligently learn about system trends by tracking file system and memory system activity patterns. The usage data that is tracked can be subsequently queried and visualized. More importantly, this data can also be mined for intelligent qualitative and quantitative system enhancements including predictive file prefetching, selective file compression and and application-driven adaptive memory allocation. We conduct an in-depth performance evaluation to demonstrate the potential benefits of the proposed system.