Verifying Causality between Distant Performance Phenomena in Large-Scale MPI Applications

Hermanns, Marc-André; Geimer, Markus; Wolf, Felix; Wylie, Brian J. N.

doi:10.1109/pdp.2009.50

Cited by 22 publications

(27 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It would therefore be desirable if the effects of redistributing a given delay could be automatically predicted and the expected savings be determined without altering the application itself. Since the effects of such changes are hard to quantify analytically, we plan to combine our delay analysis with a framework developed earlier by the authors [10], [24] that can simulate these changes via a real-time replay of event traces after they have been modified to reflect the redistributed load. First results indicate both high accuracy and good scalability, further application studies are in progress.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Identifying the Root Causes of Wait States in Large-Scale Parallel Applications

Böhme

Geimer²,

Wolf

et al. 2010

2010 39th International Conference on Parallel Processing

Self Cite

View full text Add to dashboard Cite

Driven by growing application requirements and accelerated by current trends in microprocessor design, the number of processor cores on modern supercomputers is increasing from generation to generation. However, load or communication imbalance prevents many codes from taking advantage of the available parallelism, as delays of single processes may spread wait states across the entire machine. Moreover, when employing complex point-to-point communication patterns, wait states may propagate along far-reaching cause-effect chains that are hard to track manually and that complicate an assessment of the actual costs of an imbalance. Building on earlier work by Meira Jr. et al., we present a scalable approach that identifies program wait states and attributes their costs in terms of resource waste to their original cause. By replaying event traces in parallel both in forward and backward direction, we can identify the processes and call paths responsible for the most severe imbalances even for runs with tens of thousands of processes.

show abstract

Section: Discussionmentioning

confidence: 99%

“…Furthermore, the statistical inference process may prove inaccurate for applications with highly time-dependent performance behavior [9]. Finally, the lack of traces precludes the later simulation of imbalance smoothing to narrow the space of potential optimizations, as proposed by Hermanns et al [10].…”

Section: Introductionmentioning

confidence: 99%

Identifying the Root Causes of Wait States in Large-Scale Parallel Applications

Böhme

Geimer²,

Wolf

et al. 2010

2010 39th International Conference on Parallel Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…PSiNS uses direct execution for MPI calls. SILAS (SImulation of LArgeScale parallel applications) [11] is a parallel trace-based performance simulator for large scale target systems. SILAS focuses on the effects of fine-grain alterations of applicationlevel behavior with respect to the performance under an identical execution configuration.…”

Section: Related Workmentioning

confidence: 99%

HAEC-SIM: A Simulation Framework for Highly Adaptive Energy-Efficient Computing Platforms

Bielert¹,

Ciorba²,

Feldhoff³

et al. 2015

Proceedings of the Eighth EAI International Conference on Simulation Tools and Techniques

View full text Add to dashboard Cite

This work presents a new trace-based parallel discrete event simulation framework designed for predicting the behavior of a novel computing platform running energy-aware parallel applications. Discrete event traces capture the runtime behavior of parallel applications on existing systems and form the basis for the simulation. The simulation framework processes the events of the input trace by applying simulation models that modify event properties. Thus, the output are again event traces that describe the predicted application behavior on the simulated target platform. Both input and simulated traces can be visualized and analyzed with established tools. The modular design of the framework enables the simulation of different aspects such as temporal performance and energy efficiency by applying distinct simulation models e.g.: (i) A performance model for communication that allows to evaluate the target communication topology and link properties. (ii) An energy model for computations that is based on measurements of current hardware. We showcase the potential of this simulation by simulating the execution of benchmark applications to explore design alternatives of highly adaptive and energy-efficient computing applications and platforms.

show abstract

“…This trace replay is usually much faster than direct execution, as the computation and communications are not actually executed but abstracted as trace events. A number of tools [2], [24], [40], [32], [46], [23], including SMPI, support the off-line approach.…”

Section: The Smpi Frameworkmentioning

confidence: 99%

Toward Better Simulation of MPI Applications on Ethernet/TCP Networks

Bedaride

Degomme

Genaud

et al. 2014

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract-Simulation and modeling for performance prediction and profiling is essential for developing and maintaining HPC code that is expected to scale for next-generation exascale systems, and correctly modeling network behavior is essential for creating realistic simulations. In this article we describe an implementation of a flow-based hybrid network model that accounts for factors such as network topology and contention, which are commonly ignored by other approaches. We focus on large-scale, Ethernet-connected systems, as these currently compose 37.8% of the TOP500 index, and this share is expected to increase as higher-speed 10 and 100GbE become more available. The European Mont-Blanc project to study exascale computing by developing prototype systems with low-power embedded devices will also use Ethernet-based interconnect. Our model is implemented within SMPI, an open-source MPI implementation that connects real applications to the SimGrid simulation framework. SMPI provides implementations of collective communications based on current versions of both OpenMPI and MPICH. SMPI and SimGrid also provide methods for easing the simulation of largescale systems, including shadow execution, memory folding, and support for both online and offline (i.e., post-mortem) simulation. We validate our proposed model by comparing traces produced by SMPI with those from real world experiments, as well as with those obtained using other established network models. Our study shows that SMPI has a consistently better predictive power than classical LogP-based models for a wide range of scenarios including both established HPC benchmarks and real applications.

show abstract

Verifying Causality between Distant Performance Phenomena in Large-Scale MPI Applications

Cited by 22 publications

References 12 publications

Identifying the Root Causes of Wait States in Large-Scale Parallel Applications

Identifying the Root Causes of Wait States in Large-Scale Parallel Applications

HAEC-SIM: A Simulation Framework for Highly Adaptive Energy-Efficient Computing Platforms

Toward Better Simulation of MPI Applications on Ethernet/TCP Networks

Contact Info

Product

Resources

About