Improving parallel job scheduling by combining gang scheduling and backfilling techniques

Zhang, Y.; Franke, Hubertus; Moreira, José E.; Sivasubramaniam, Anand

doi:10.1109/ipdps.2000.845975

Cited by 67 publications

(58 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…refer to [8,43,44,10,12,16,18,32,33,34] and the references therein). At the same time, scheduling can be used to improve the fault-tolerance [1,27] of a system in three broad ways.…”

Section: Related Workmentioning

confidence: 99%

Performance Implications of Failures in Large-Scale Cluster Scheduling

Zhang

Squillante

Sivasubramaniam

et al. 2005

Job Scheduling Strategies for Parallel Processing

View full text Add to dashboard Cite

Abstract. As we continue to evolve into large-scale parallel systems, many of them employing hundreds of computing engines to take on mission-critical roles, it is crucial to design those systems anticipating and accommodating the occurrence of failures. Failures become a commonplace feature of such large-scale systems, and one cannot continue to treat them as an exception. Despite the current and increasing importance of failures in these systems, our understanding of the performance impact of these critical issues on parallel computing environments is extremely limited. In this paper we develop a general failure modeling framework based on recent results from large-scale clusters and then we exploit this framework to conduct a detailed performance analysis of the impact of failures on system performance for a wide range of scheduling policies. Our results demonstrate that such failures can have a significant impact on the mean job response time and mean job slowdown under existing scheduling policies that ignore failures. We therefore investigate different scheduling mechanisms and policies to address these performance issues. Our results show that periodic checkpointing of jobs seems to do little to ease this problem. On the other hand, we demonstrate that information about the spatial and temporal correlation of failure occurrences can be very useful in designing a scheduling (job allocation) strategy to enhance system performance, with the former providing the greatest benefits.

show abstract

“…refer to [8,43,44,10,12,16,18,32,33,34] and the references therein). At the same time, scheduling can be used to improve the fault-tolerance [1,27] of a system in three broad ways.…”

Section: Related Workmentioning

confidence: 99%

Performance Implications of Failures in Large-Scale Cluster Scheduling

Zhang

Squillante

Sivasubramaniam

et al. 2005

Job Scheduling Strategies for Parallel Processing

View full text Add to dashboard Cite

show abstract

“…Further research [2] comparing gang scheduling with batch scheduling has shown that augmenting gang scheduling with backfill provides additional performance benefits under time-sharing. Another simulation study [3] comparing gang scheduling to batch scheduling also shows performance benefits for time-sharing in certain cases.…”

Section: Related Workmentioning

confidence: 99%

Time-Sharing Redux for Large-Scale HPC Systems

Hofmeyr

Iancu

Colmenares

et al. 2016

2016 IEEE 18th International Conference on High Performance Computing and Communications; IEEE 14th International Conference On

View full text Add to dashboard Cite

Abstract-HPC facilities typically use batch scheduling to space-share jobs. In this paper we revisit time-sharing using a trace of 2.4 million jobs obtained during 20 months of operation of a modern petascale supercomputer. Our simulations show that batch scheduling produces distributions that are skewed towards smaller jobs and longer execution times, whereas timesharing produces much more uniform slowdowns. Consequently, for applications that strong scale, the turnaround time does not scale with batch scheduling, but it does with time-sharing, resulting in turnarounds that are orders of magnitude better at the largest scales. In addition, we show that time-sharing can confer additional benefits with modern programming practices and in noisy systems. FutureExascale HPC systems, are expected to exhibit billion-way heterogeneous parallelism and poor performance predictability. As many applications will run in strong scaling, how resource allocation policies affect the experience of supercomputer users has once again become a timely subject.

show abstract

“…This problem is accentuated when we multiprogram a large number of tasks which can involve considerable swapping overheads. For this reason, several systems (e.g., [29,32] ) limit the multiprogramming level (MPL) so that swapping overheads are kept to a minimum. Under the same rationale, we use the multiprogramming level directly to modulate the memory usage rather than explicitly model the memory consumption.…”

Section: -Cpu Schedulermentioning

confidence: 99%

“…Further, the tasks from the same job must be scheduled into the same row, which makes it more challenging. We have implemented the exact co-scheduling heuristics that are proposed in our earlier work [30,29,28,33,32].…”

Section: Implementing the Front Endmentioning

confidence: 99%

ClusterSchedSim: A Unifying Simulation Framework for Cluster Scheduling Strategies

Zhang

Sivasubramaniam

2004

SIMULATION

View full text Add to dashboard Cite

As clusters are being deployed to support a wide range of parallel workloads, scheduling becomes a challenging research issue because these workloads exhibit diverse characteristics and impose varying quality-of-service requirements. Many scheduling strategies are thus proposed, each intended for a different application/system setting. Due to the lack of a uniform simulation platform, a significant amount of research effort is spent in building a unique simulator for each algorithm, which may lead to false conclusions. This article presents ClusterSchedSim, which is a unifying simulation frame-work of cluster scheduling strategies. The core of ClusterSchedSim includes the node model and an interconnect model. ClusterSchedSim has implemented variations of popular cluster scheduling schemes, and it is flexible enough to add on new schemes. Using ClusterSchedSim, one can conveniently compare different scheduling schemes, profile their executions, and understand the impact of different application and system configuration parameters.

show abstract

Improving parallel job scheduling by combining gang scheduling and backfilling techniques

Cited by 67 publications

References 9 publications

Performance Implications of Failures in Large-Scale Cluster Scheduling

Performance Implications of Failures in Large-Scale Cluster Scheduling

Time-Sharing Redux for Large-Scale HPC Systems

ClusterSchedSim: A Unifying Simulation Framework for Cluster Scheduling Strategies

Contact Info

Product

Resources

About