Using perturbation tracking to compensate for intrusion in message-passing systems

Gannon, J.A.; Williams, K. J. Orton; Andersland, Mark S.; Lummp, J.E.; Gasavant, T.L.

doi:10.1109/icdcs.1994.302445

Cited by 9 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…First, instrumentation is necessary in either the application itself or a library that the application calls. Second, the perturbation introduced by tracing can change the results of the analysis [7]. Third, tracing generates a tremendous volume of data.…”

Section: Tracing Mpi Applicationsmentioning

confidence: 99%

Performance analysis of distributed applications using automatic classification of communication inefficiencies

Vetter

2000

Proceedings of the 14th International Conference on Supercomputing

View full text Add to dashboard Cite

We present a technique for performance analysis that helps users understand the communication behavior of their message passing applications. Our method automatically classifies individual communication operations and it reveals the cause of communication inefficiencies in the application. This classification allows the developer to focus quickly on the culprits of truly inefficient behavior, rather than manually foraging through massive amounts of performance data. Specifically, we trace the message operations of MPI applications and then classify each individual communication event using decision tree classification, a supervised learning technique. We train our decision tree using microbenchmarks that demonstrate both efficient and inefficient communication. Since our technique adapts to the target system's configuration through these microbenchmarks, we can simultaneously automate the performance analysis process and improve classification accuracy. Our experiments on four applications demonstrate that our technique can improve the accuracy of performance analysis, and dramatically reduce the amount of data that users must encounter.

show abstract

Section: Tracing Mpi Applicationsmentioning

confidence: 99%

Performance analysis of distributed applications using automatic classification of communication inefficiencies

Vetter

2000

Proceedings of the 14th International Conference on Supercomputing

View full text Add to dashboard Cite

show abstract

“…Several researchers have developed techniques to attempt to remove overheads from the reported data [14,20,69,72,76]. Yan and Listgarten [76] specifically addressed the overhead of writing the trace buffer to disk in AIMS by generating an event marker for these write operations and removing the overhead in a postprocessing step.…”

Section: Perturbationmentioning

confidence: 99%

“…If we choose a threshold of 0.2, then the highest the computed distance can be for a match is 10.2, so s2 and sl will not match using any of the Minkowski distances. When we compare sO, (50,1,20,21,49), with s2, we get distances of 8, 4.5, and 3. The maximum value in the two vectors is 50, so the highest the distances can be for a match is 10.…”

Section: Minkowski Distancementioning

confidence: 99%

Scalable event tracking on high-end parallel systems

Mohror¹

2000

View full text Add to dashboard Cite

Accurate performance analysis of high end systems requires event-based traces to correctly identify the root cause of a number of the complex performance problems that arise on these highly parallel systems. These high-end architectures contain tens to hundreds of thousands of processors, pushing application scalability challenges to new heights. Unfortunately, the collection of event-based data presents scalability challenges itself: the large volume of collected data increases tool overhead, and results in data files that are difficult to store and analyze. Our solution to these problems is a new measurement technique called trace profiling that collects the information needed to diagnose performance problems that traditionally require traces, but at a greatly reduced data volume. The trace profiling technique reduces the amoun! of data measured and stored by capitalizing on the repeated behavior of programs, and on the similarity of the behavior and performance of parallel processes in an application run. Trace profiling is a hybrid between profiling and tracing, collecting summary information about the event patterns in an application run. Because the data has already been classified into behavior categories, we can present reduced, partially analyzed performance data to the user, highlighting the performance behaviors that comprised most of the execution time.

show abstract

“…Postprocessing of the traces is carried out to compensate for intrusive effects of monitoring. Such techniques have been explored by Malony et al [7,91 and Casavant et al [4]. Models of perturbation in execution have been developed and used to process traces for predicting true performance.…”

Section: Introductionmentioning

confidence: 99%

Dynamic techniques for minimizing the intrusive effect of monitoring actions

Gupta

Spezialetti²

Proceedings of 15th International Conference on Distributed Computing Systems

View full text Add to dashboard Cite

Programs written in distributed programming lanuages are often non-deterministac an nature and ience the execution of such programs is not only dependent upon the rogram znput, but also on the tamzng of the execution. Fhus, an attempt t o monitor the runtime behavior of a non-determinist@ program through code instrumentation, such as durang debugging, can potentially alter the program's behavior. In this paper we present techni ues for dynamzcally minimizing the intrusive e ects of monitoring b y attempting t o ensure deterministic events are the same for uninstrumented and instrumented programs. This goal is achieved b y minimizing the intrusave effects of monitoring on message passing between processes located at different sites of a distributed system. Monitorang aclaons alter the message pool at the time of non-deterministic message selection and the order in which the messages arrive at a processor. The dynamic techniques presented in this paper minimize intrusion by restoring the message pool and message orderang.that the li f elihoods of various outcomes of the non-

show abstract

Using perturbation tracking to compensate for intrusion in message-passing systems

Cited by 9 publications

References 12 publications

Performance analysis of distributed applications using automatic classification of communication inefficiencies

Performance analysis of distributed applications using automatic classification of communication inefficiencies

Scalable event tracking on high-end parallel systems

Dynamic techniques for minimizing the intrusive effect of monitoring actions

Contact Info

Product

Resources

About