Performance data collection using a hybrid approach

Metz, Edu; Lencevicius, Raimondas; Gonzalez, Teofilo F.

doi:10.1145/1081706.1081729

Cited by 18 publications

(11 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Combined with tracing, it can give performance engineers a sufficiently detailed analysis of the system with low overhead [46], such as event relationships in time and reconstructing the dynamic behavior of a software system. In addition, sampling-based execution monitoring can be applied to code coverage testing [61] which finds the code exercised by a particular set of test input.…”

Section: Thesis Contributionmentioning

confidence: 99%

Sampling-based program execution monitoring

Fischmeister

2010

Proceedings of the ACM SIGPLAN/SIGBED 2010 Conference on Languages, Compilers, and Tools for Embedded Systems

View full text Add to dashboard Cite

For its high overall cost during product development, program debugging is an important aspect of system development. Debugging is a hard and complex activity, especially in time-sensitive systems which have limited resources and demanding timing constraints.System tracing is a frequently used technique for debugging embedded systems. A specific use of system tracing is to monitor and debug control-flow problems in programs. However, it is difficult to implement because of the potentially high overhead it might introduce to the system and the changes which can occur to the system behaviour due to tracing.To solve the above problems, in this work, we present a sampling-based approach to program execution monitoring which specifically helps developers trace the program execution in time-sensitive systems such as real-time applications. We build the system model and propose three theorems which determine the sampling period or the optimal in different scenarios. We also design seven heuristics and an instrumentation framework to extend the sampling period which can reduce the monitoring overhead and achieve an optimal tradeoff between accuracy and overhead introduced by instrumentation. Using this monitoring framework, we can use the information extracted through sampling to reconstruct the system state and execution paths to locate the deviation. Based on the statistically significant data, we also model the trend of the sampling period with the instrumentation steps. Based on the modelling results, we devise a scheme for predicting the number of markers we need to reach a certain sampling period. Last, we build a tool chain to instrument and monitoring the software system and further prove the soundness of our approach.

show abstract

Section: Thesis Contributionmentioning

confidence: 99%

Sampling-based program execution monitoring

Fischmeister

2010

Proceedings of the ACM SIGPLAN/SIGBED 2010 Conference on Languages, Compilers, and Tools for Embedded Systems

View full text Add to dashboard Cite

show abstract

“…Sometimes, such as with debugger breakpoints, the instrumentation is hardware assisted to make it much more efficient. Some monitoring approaches may use timer-based instrumentation, such as sampling-based code profiling (e.g., [16]), while others may use memory page protections to signal an event (e.g., [21]), and still others use probabilistic or counter-based sampling (e.g., [1,15]). Effectively creating the necessary instrumentation, and making it efficient, are complex and technologically difficult tasks, often involving creating specialized instrumentation tools by hand.…”

Section: Runtime Monitoringmentioning

confidence: 99%

Using AOP for detailed runtime monitoring instrumentation

Nusayr

Cook

2009

Proceedings of the Seventh International Workshop on Dynamic Analysis

View full text Add to dashboard Cite

Although AOP has long been used for monitoring purposes, the level of detail that virtually all AOP frameworks support is not enough to support broad classes of monitoring applications. Often the method or function call is the basic unit of code weaving, and if some data access weaving is supported it is usually very limited, e.g., to object field access. In this paper we demonstrate the need for AOP to be extended if it is to support broad runtime monitoring needs, and then present two new joinpoint types for AspectJ, the basic block and loop back edge types. These allow much more fine-grained weaving of advice, which in turn supports a much larger category of runtime monitoring applications. We demonstrate the feasibility of these extensions with example prototype implementations using the AspectBench Compiler (abc).

show abstract

“…More instrumentation codes can facilitate debugging and optimizing, but they result in additional distortions, which inhibit the insertion of as many instrumentation codes as required. Thus, there have been studies to reduce the overhead of the instrumentation (Metz et al, 2003). Hardware-assisted measurement does not cause distortion, but its precision and scope are limited (Stewart and Arora, 2003).…”

Section: Introductionmentioning

confidence: 99%