Task Bench: A Parameterized Benchmark for Evaluating Parallel Runtime Performance

Slaughter, Elliott; Wu, Wei; Fu, Yuankun; Brandenburg, Legend; Garcia, Nicolai; Kautz, Wilhem; Marx, Emily; Morris, Kaleb S.; Cao, Qinglei; Bosilca, George; Mirchandaney, Seema; Leek, Wonchan; Treichlerk, Sean; McCormick, Patrick; Aiken, Alex

doi:10.1109/sc41405.2020.00066

Cited by 43 publications

(31 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Even still, this is comparable to the smallest METG observed in Ref. [2] for the Spark, Dask, or Swift/T workflow systems. dwork has an overhead of 1 second at the largest node counts, comparable or smaller to that measured in Ref.…”

Section: Discussionsupporting

confidence: 62%

“…This is slightly above the fastest schedulers from Ref. [2] because the synchronization is global, and not point-to-point. The METG for mpi-list is the largest, around 1-5 seconds.…”

Section: Discussionmentioning

confidence: 86%

“…The lower portion of Figure 4 reproduces the minimum effective task granularity (METG) plot of Ref. [2]. Computational efficiency, plotted on the vertical axis, is defined as ideal divided by actual per-task execution time.…”

Section: Scaling Resultsmentioning

confidence: 99%

“…I quantify the overhead of each task scheduling tool in order to understand the 'minimum effective task granularity' (METG). This measure was introduced by Ref [2] to measure the overhead incurred during actual task processing. Basically, it measures (in seconds) the task difficulty needed to equally divide total execution time between scheduling overhead and actual work done on the task.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Rogers¹

2021

Preprint

View full text Add to dashboard Cite

Runtime scheduling and workflow systems are an increasingly popular algorithmic component in HPC because they allow full system utilization with relaxed synchronization requirements. There are so many special-purpose tools for task scheduling, one might wonder why more are needed. Use cases seen on the Summit supercomputer needed better integration with MPI and greater flexibility in job launch configurations. Preparation, execution, and analysis of computational chemistry simulations at the scale of tens of thousands of processors revealed three distinct workflow patterns. A separate job scheduler was implemented for each one using extremely simple and robust designs: file-based, task-list based, and bulk-synchronous. Comparing to existing methods shows unique benefits of this work, including simplicity of design, suitability for HPC centers, short startup time, and well-understood per-task overhead. All three new tools have been shown to scale to full utilization of Summit, and have been made publicly available with tests and documentation. This work presents a complete characterization of the minimum effective task granularity for efficient scheduler usage scenarios. These schedulers have the same bottlenecks, and hence similar task granularities as those reported for existing tools following comparable paradigms.

show abstract

Section: Discussionsupporting

confidence: 62%

“…This is slightly above the fastest schedulers from Ref. [2] because the synchronization is global, and not point-to-point. The METG for mpi-list is the largest, around 1-5 seconds.…”

Section: Discussionmentioning

confidence: 86%

Section: Scaling Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Rogers¹

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Tile size is also a critical parameter for tile-based algorithm, trading off between performance per task and concurrency between tasks. A more in depth analysis of the performance/concurrency tradeoff for regular problems (where the cost of all tasks is similar) can be found in [33] for multiple runtime systems. For TLR Cholesky, [23] proposed a model to calculate the approximate optimal tile size with the assuming a first-order approximation -the serial part (the critical path in the algorithm) at distance one overlapping with the parallel part (everything outside the critical path).…”

Section: Suitable Tile Size Selectionmentioning

confidence: 99%

Leveraging PaRSEC Runtime Support to Tackle Challenging 3D Data-Sparse Matrix Problems

Cao

Pei

Akbudak

et al. 2021

2021 IEEE International Parallel and Distributed Processing Symposium (IPDPS)

Self Cite

View full text Add to dashboard Cite

The task-based programming model associated with dynamic runtime systems has gained popularity for challenging problems because of workload imbalance, heterogeneous resources, or extreme concurrency. During the last decade, lowrank matrix approximations, where the main idea consists of exploiting data sparsity typically by compressing off-diagonal tiles up to an application-specific accuracy threshold, have been adopted to address the curse of dimensionality at extreme scale. In this paper, we create a bridge between the runtime and the linear algebra by communicating knowledge of the data sparsity to the runtime. We design and implement this synergistic approach with high user productivity in mind, in the context of the PaRSEC runtime system and the HiCMA numerical library. This requires to extend PaRSEC with new features to integrate rank information into the dataflow so that proper decisions can be taken at runtime. We focus on the tile low-rank (TLR) Cholesky factorization for solving 3D data-sparse covariance matrix problems arising in environmental applications. In particular, we employ the 3D exponential model of Mateŕn matrix kernel, which exhibits challenging nonuniform high ranks in off-diagonal tiles. We first provide a dynamic data structure management driven by a performance model to reduce extra floating-point operations. Next, we optimize the memory footprint of the application by relying on a dynamic memory allocator, and supported by a rank-aware data distribution to cope with the workload imbalance. Finally, we expose further parallelism using kernel recursive formulations to shorten the critical path. Our resulting high-performance implementation outperforms existing data-sparse TLR Cholesky factorization by up to 7-fold on a large-scale distributed-memory system, while minimizing the memory footprint up to a 44-fold factor. This multidisciplinary work highlights the need to empower runtime systems beyond their original duty of task scheduling for servicing next-generation low-rank matrix algebra libraries.

show abstract

Performance comparison of Dask and Apache Spark on HPC systems for neuroimaging

Dugré

Hayot‐Sasson

Glatard

2023

Concurrency and Computation

View full text Add to dashboard Cite

SummaryThe general increase in data size and data sharing motivates the adoption of Big Data strategies in several scientific disciplines. However, while several options are available, no particular guidelines exist for selecting a Big Data engine. In this paper, we compare the runtime performance of two popular Big Data engines with Python APIs, Apache Spark, and Dask, in processing neuroimaging pipelines. Our experiments use three synthetic neuroimaging applications to process the 606 GB BigBrain image and an actual pipeline to process data from thousands of anatomical images. We benchmark these applications on a dedicated HPC cluster running the Lustre file system while using varying combinations of the number of nodes, file size, and task duration. Our results show that although there are slight differences between Dask and Spark, the performance of the engines is comparable for data‐intensive applications. However, Spark requires more memory than Dask, which can lead to slower runtime depending on configuration and infrastructure. In general, the limiting factor was the data transfer time. While both engines are suitable for neuroimaging, more efforts need to be put to reduce the data transfer time and the memory footprint of applications.

show abstract

Task Bench: A Parameterized Benchmark for Evaluating Parallel Runtime Performance

Cited by 43 publications

References 24 publications

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Leveraging PaRSEC Runtime Support to Tackle Challenging 3D Data-Sparse Matrix Problems

Performance comparison of Dask and Apache Spark on HPC systems for neuroimaging

Contact Info

Product

Resources

About