Because today's software is increasingly dynamic and runs in heterogeneous environments, it is difficult to assess software systems outside the actual context in which they execute. This situation has spurred much interest in efficient runtime monitoring of software systems' dynamic behavior in production environments. Most existing runtime monitoring approaches, however, suffer from three limitations: they are ad-hoc and hard to extend/adapt; they impose too much overhead to be practical; and they are defined within only one computational layer (hardware, O/S, VM, or software). This project will investigate and define a general, efficient runtime monitoring approach that leverages software and hardware capabilities in a synergistic way. In this approach, monitoring tasks will be specified using a language that provides suitable abstractions. These specifications will then be automatically analyzed and the monitoring tasks distributed across software and hardware layers to minimize performance overhead. This project will improve the state of the art by (1) providing hardware and software support for new monitoring activities currently neglected as impractical, (2) developing infrastructure and tools that will be made available to the research community, and (3) improving overall software quality through detection of and reaction to anomalous, erratic, or malicious software behavior.
