Log-based architectures for general-purpose monitoring of deployed code

Flexible Hardware Acceleration for Instruction-Grain Lifeguards

“…Other studies have proposed more general-purpose hardware solutions. For example, our previous study proposed log-based architectures (LBA) 12 , which captures a log from a monitored program and ships it to another on-chip core that executes the monitoring functionality (see the ''Log-Based Architectures'' sidebar for more details on this approach). An earlier study 13 proposed dynamic instruction stream editing (DISE), which performs patternmatching based dynamic rewriting of a processor's instruction stream to insert calls to monitoring code.…”

“…Both DISE and our earlier work on LBA focus only on reducing the costs of DBI, such as reducing the resource competition between monitored programs and lifeguards. As a result, the instructiongrain monitoring overhead, although significantly reduced, is still large (for example, 3 to 5Â slowdowns 12 ). We believe that the key to long-term impact is fast and flexible lifeguards, for several reasons.…”

Flexible Hardware Acceleration for Instruction-Grain Program Monitoring

“…Unfortunately, these mechanisms are useful only for the narrow class of lifeguards that they support. Other studies [3,5] have proposed more general-purpose hardware solutions. For example, our previous study proposed Log-Based Architectures (LBA) [3], which capture a log from a monitored program and ship it to another on-chip core that executes the monitoring functionality.…”

PaRV: Parallelizing Runtime Detection and Prevention of Concurrency Errors

“…ParaLog [8] extends work on log-based architectures [3] provide hardware support for instrumenting, logging and monitoring executions of multithreaded programs. Techniques in ParaLog not only reduce the application slowdown due to instrumentation and logging, but also allow, similarly to PaRV, parallelized monitoring algorithms to be run on separate resources from the application, thus further reducing slowdown.…”

“…A related approach is that of log-based architectures (e.g., [3], [6] ) which provide on-chip hardware resources for reducing the runtime overhead that comes from monitoring executions. Differently from log-based approaches, our tool does not need any additional hardware support but can benefit from it.…”