Toward Better Simulation of MPI Applications on Ethernet/TCP Networks

Bedaride, Paul; Degomme, Augustin; Genaud, Stéphane; Legrand, Arnaud; Markomanolis, George S.; Quinson, Martin; Stillwell, Mark; Suter, Frédéric; Videau, Brice

doi:10.1007/978-3-319-10214-6_8

Cited by 25 publications

(30 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Thus, we uphold our resource requirement goal. (2) Second, SMPI builds on the hybrid flow‐level network models of SimGrid that allow to faithfully model communications and contention, which is essential in the context of our study since load balancing in AMPI induces Virtual Process migration that can be costly in terms of communications.…”

Section: Sampi: a Simulation Workflow To Evaluate Ampi Performancementioning

confidence: 99%

Performance modeling of a geophysics application to accelerate over‐decomposition parameter tuning through simulation

Tesser

Schnorr

Legrand

et al. 2018

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

Summary Finite‐difference methods are commonplace in High Performance Computing applications. Despite their apparent regularity, they often exhibit load imbalance that damages their efficiency. We characterize the spatial and temporal load imbalance of Ondes3D, a typical finite‐differences application dedicated to earthquake modeling. Our analysis reveals imbalance originating from the structure of the input data, and from low‐level CPU optimizations. Ondes3D was successfully ported to AMPI/CHARM++ using over‐decomposition and MPI process migration techniques to dynamically rebalance the load. However, this approach requires careful selection of the over‐decomposition level, the load balancing algorithm, and its activation frequency. These choices are usually tied to application structure and platform characteristics. In this article, we propose a workflow that leverages the capabilities of SimGrid to conduct such study at low experimental cost. We rely on a combination of emulation, simulation, and application modeling that requires minimal code modification and manages to capture both spatial and temporal load imbalance to faithfully predict the performance of dynamic load balancing. We evaluate the quality of our simulation by comparing simulation results with the outcome of real executions and demonstrate how this approach can be used to quickly find the optimal load balancing configuration for a given application/hardware configuration.

show abstract

Section: Sampi: a Simulation Workflow To Evaluate Ampi Performancementioning

confidence: 99%

Performance modeling of a geophysics application to accelerate over‐decomposition parameter tuning through simulation

Tesser

Schnorr

Legrand

et al. 2018

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the HPC context, more precisely regarding MPI applications, SimGrid provides SMPI module [8] which can be used in two different modes: offline and online. In the offline mode, a trace of a previous native execution on a supercomputer is replayed in the simulator.…”

Section: Simgrid Simulation Frameworkmentioning

confidence: 99%

Platform Independent Profiling of a QCD Code

Marinkovic¹,

Stanisic²

2017

Proceedings of 34th Annual International Symposium on Lattice Field Theory — PoS(LATTICE2016)

View full text Add to dashboard Cite

The supercomputing platforms available for high performance computing based research evolve at a great rate. However, this rapid development of novel technologies requires constant adaptations and optimizations of the existing codes for each new machine architecture. In such context, minimizing time of efficiently porting the code on a new platform is of crucial importance. A possible solution for this common challenge is to use simulations of the application that can assist in detecting performance bottlenecks. Due to prohibitive costs of classical cycle-accurate simulators, coarse-grain simulations are more suitable for large parallel and distributed systems. We present a procedure of implementing the profiling for openQCD code [1] through simulation, which will enable the global reduction of the cost of profiling and optimizing this code commonly used in the lattice QCD community. Our approach is based on well-known SimGrid simulator [2], which allows for fast and accurate performance predictions of HPC codes. Additionally, accurate estimations of the program behavior on some future machines, not yet accessible to us, are anticipated. 34th annual International Symposium on Lattice Field Theory

show abstract

“…In this work we have used the SimGrid simulation framework to model, control, reproduce, and monitor large-scale computing systems to analyze the performance of the DLS algorithms on computationally intensive scientific applications. We have chosen SimGrid to implement the DLS algorithms and to evaluate their robustness because it is a versatile tool and is also validated for the evaluation of scheduling algorithms in distributed and heterogeneous computing systems, such as grids, clouds or peer-to-peer systems [27] [28] [29]. SimGrid's deployment and platform files facilitate the creation of the execution scenarios reflecting load imbalance for performing an experimental evaluation of the robustness of the DLS algorithms.…”

Section: Problemmentioning

confidence: 99%

Portfolio-Based Selection of Robust Dynamic Loop Scheduling Algorithms Using Machine Learning

Sukhija

Malone

Srivastava

et al. 2014

2014 IEEE International Parallel &Amp; Distributed Processing Symposium Workshops

View full text Add to dashboard Cite

The execution of computationally intensive parallel applications in heterogeneous environments, where the quality and quantity of computing resources available to a single user continuously change, often leads to irregular behavior, in general due to variations of algorithmic and systemic nature. To improve the performance of scientific applications, loop scheduling algorithms are often employed for load balancing of their parallel loops. However, it is a challenge to select the most robust scheduling algorithms for guaranteeing optimized performance of scientific applications on large-scale computing systems that comprise resources which are widely distributed, highly heterogeneous, often shared among multiple users, and have computing availabilities that cannot always be guaranteed or predicted. To address this challenge, in this work we focus on a portfolio-based approach to enable the dynamic selection and use of the most robust dynamic loop scheduling (DLS) algorithm from a portfolio of DLS algorithms, depending on the given application and current system characteristics including workload conditions. Thus, in this paper we provide a solution to the algorithm selection problem and experimentally evaluate its quality. We propose the use of supervised machine learning techniques to build empirical robustness prediction models that are used to predict DLS algorithm's robustness for given scientific application characteristics and system availabilities. Using simulated scientific applications characteristics and system availabilities, along with empirical robustness prediction models, we show that the proposed portfolio-based approach enables the selection of the most robust DLS algorithm that satisfies a user-specified tolerance on the given application's performance obtained in the particular computing system with a certain variable availability. We also show that the portfoliobased approach offers higher guarantees regarding the robust performance of the application using the automatically selected DLS algorithms when compared to the robust performance of the same application using a manually selected DLS algorithm.

show abstract

Toward Better Simulation of MPI Applications on Ethernet/TCP Networks

Cited by 25 publications

References 39 publications

Performance modeling of a geophysics application to accelerate over‐decomposition parameter tuning through simulation

Performance modeling of a geophysics application to accelerate over‐decomposition parameter tuning through simulation

Platform Independent Profiling of a QCD Code

Portfolio-Based Selection of Robust Dynamic Loop Scheduling Algorithms Using Machine Learning

Contact Info

Product

Resources

About