Hints to improve automatic load balancing with LeWI for hybrid applications

García, Marta; Labarta, Jesús; Corbalan, Julita

doi:10.1016/j.jpdc.2014.05.004

Cited by 34 publications

(32 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This observation suggests that using a predictive method for job's runtime, i.e. based on machine learning, rather than asking the user, [16] 30.5% small to high small/med high low STREAM [27] 30.8% small to high small/med low high CoreNeuron [21] 35,5% small to high small to high high med NEST [22] 2.6% small to high small to high high med Alya [10] 0.6% small high high med…”

Section: Evaluation Of Max_slowdownmentioning

confidence: 99%

Holistic Slowdown Driven Scheduling and Resource Management for Malleable Jobs

D'Amico

Jokanović

Corbalan

2019

Proceedings of the 48th International Conference on Parallel Processing

Self Cite

View full text Add to dashboard Cite

In job scheduling, the concept of malleability has been explored since many years ago. Research shows that malleability improves system performance, but its utilization in HPC never became widespread. The causes are the difficulty in developing malleable applications, and the lack of support and integration of the different layers of the HPC software stack. However, in the last years, malleability in job scheduling is becoming more critical because of the increasing complexity of hardware and workloads. In this context, using nodes in an exclusive mode is not always the most efficient solution as in traditional HPC jobs, where applications were highly tuned for static allocations, but offering zero flexibility to dynamic executions. This paper proposes a new holistic, dynamic job scheduling policy, Slowdown Driven (SD-Policy), which exploits the malleability of applications as the key technology to reduce the average slowdown and response time of jobs. SD-Policy is based on backfill and node sharing. It applies malleability to running jobs to make room for jobs that will run with a reduced set of resources, only when the estimated slowdown improves over the static approach. We implemented SD-Policy in SLURM and evaluated it in a real production environment, and with a simulator using workloads of up to 198K jobs. Results show better resource utilization with the reduction of makespan, response time, slowdown, and energy consumption, up to respectively 7%, 50%, 70%, and 6%, for the evaluated workloads. CCS CONCEPTS• Software and its engineering → Scheduling; • Computer systems organization → Multicore architectures.

show abstract

Section: Evaluation Of Max_slowdownmentioning

confidence: 99%

Holistic Slowdown Driven Scheduling and Resource Management for Malleable Jobs

D'Amico

Jokanović

Corbalan

2019

Proceedings of the 48th International Conference on Parallel Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…We have modified the code to add calls to poll_DROM in safe points for malleability [4]. • Pils [18] is a synthetic benchmark, doing computation-intensive operations. It is parallelized with MPI + OmpSs.…”

Section: Evaluation Of Drom-enabled System's Performancementioning

confidence: 99%

“…We propose to provide resource managers with more tools that will give them a dynamic control of resources allocated to the application and a particular feedback about the utilization of these resources. In this paper we will extend the DLB [17] [18] library with a new API designed to be used by the resource managers. This new API will offer a transversal layer in the HPC software stack to coordinate the resource manager and the parallel runtime.…”

Section: Introductionmentioning

confidence: 99%

Drom

D'Amico

García-Gasulla

López

et al. 2018

Proceedings of the 47th International Conference on Parallel Processing Companion

Self Cite

View full text Add to dashboard Cite

In the design of future HPC systems, research in resource management is showing an increasing interest in a more dynamic control of the available resources. It has been proven that enabling the jobs to change the number of computing resources at run time, i.e. their malleability, can significantly improve HPC system performance. However, job schedulers and applications typically do not support malleability due to the common belief that it introduces additional programming complexity and performance impact. This paper presents DROM, an interface that provides efficient malleability with no effort for program developers. The running application is enabled to adapt the number of threads to the number of assigned computing resources in a completely transparent way to the user through the integration of DROM with standard programming models, such as OpenMP/OmpSs, and MPI. We designed the APIs to be easily used by any programming model, application and job scheduler or resource manager. Our experimental results from two realistic use cases analysis, based on malleability by reducing the number of cores a job is using per node and jobs co-allocation, show the potential of DROM for improving the performance of HPC systems. In particular, the workload of two MPI+OpenMP neuro-simulators are tested, reporting improvement in system metrics, such as total run time and average response time, up to 8% and 48%, respectively. CCS CONCEPTS• Software and its engineering → Software libraries and repositories;© 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

show abstract

“…The students noticed a significant load imbalance while running with the hybrid MPI+OpenMP version of the application. Consequently, they have been exposed to the problem of load balancing, and they tested Dynamic Load Balancing (DLB) [30], [31], a library developed at BSC that dynamically reschedules workload at runtime to maximize resource utilization.…”

Section: A Hpc Fundamentalsmentioning

confidence: 99%

Filling the gap between education and industry: evidence-based methods for introducing undergraduate students to HPC

Banchelli

Mantovani

2018

2018 IEEE/ACM Workshop on Education for High-Performance Computing (EduHPC)

View full text Add to dashboard Cite

Educational institutions provide in most cases basic theoretical background covering several computational science topics. However, High-Performance Computing (HPC) and Parallel and Distributed Computing (PDC) markets require specialized technical profiles. Even the most skilled students are often not prepared to face production HPC applications of thousands of lines nor complex computational frameworks from other disciplines nor heterogeneous multinode machines accessed by hundreds of users. In this paper, we offer an educational package for filling this gap. Leveraging the 4-years experience of the Student Cluster Competition, we present our educational journey together with the lessons learned and the outcomes of our methodology. We show how, in a time span of a semester and an affordable budget, a university can implement an educational package preparing pupils for starting competitive professional careers. Our findings also highlight that 78% of the students exposed to our methods remain within the HPC high-education, research or industry.

show abstract

Hints to improve automatic load balancing with LeWI for hybrid applications

Cited by 34 publications

References 16 publications

Holistic Slowdown Driven Scheduling and Resource Management for Malleable Jobs

Holistic Slowdown Driven Scheduling and Resource Management for Malleable Jobs

Drom

Filling the gap between education and industry: evidence-based methods for introducing undergraduate students to HPC

Contact Info

Product

Resources

About