Using Moldability to Improve the Performance of Supercomputer Jobs

Cirne, Walfredo; Berman, Francine

doi:10.1006/jpdc.2002.1869

Cited by 35 publications

(26 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The cost of the resources however is not taken into consideration. In a number of earlier works [29,30] the size of the resource request is optimized so that the sum of the wait time to get the requested resource and the run time of the application on the resource is minimized. With the advent of cloud computing that provides resources on demand these issues become irrelevant as there is no wait time involved to get the resources.…”

Section: Related Workmentioning

confidence: 99%

The cost of doing science on the cloud: The Montage example

Deelman¹,

Singh²,

Livny³

et al. 2008

2008 SC - International Conference for High Performance Computing, Networking, Storage and Analysis

513

389

View full text Add to dashboard Cite

show abstract

Section: Related Workmentioning

confidence: 99%

The cost of doing science on the cloud: The Montage example

Deelman¹,

Singh²,

Livny³

et al. 2008

2008 SC - International Conference for High Performance Computing, Networking, Storage and Analysis

513

389

View full text Add to dashboard Cite

show abstract

“…There have been some studies addressing a moldable job scheduling model. Cirne [14] proposed a moldable job scheduling scheme in which an application level scheduler makes this choice for the user before job submission. Our idea is similar to this, but the difference is that our system does not rely on a set of possible requests from job users but makes molding decisions on its own within a certain limitation.…”

Section: Job Schedulingmentioning

confidence: 99%

Job Mapping and Scheduling on Free-Space Optical Networks

Fujiwara

Koibuchi

2016

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYA number of parallel applications run on a highperformance computing (HPC) system simultaneously. Job mapping and scheduling become crucial to improve system utilization, because fragmentation prevents an incoming job from being assigned even if there are enough compute nodes unused. Wireless supercomputers and datacenters with free-space optical (FSO) terminals have been proposed to replace the conventional wired interconnection so that a diverse application workload can be better supported by changing their network topologies. In this study we firstly present an efficient job mapping by swapping the endpoints of FSO links in a wireless HPC system. Our evaluation shows that an FSOequipped wireless HPC system can achieve shorter average queuing length and queuing time for all the dispatched user jobs. Secondly, we consider the use of a more complicated and enhanced scheduling algorithm, which can further improve the system utilization over different host networks, as well as the average response time for all the dispatched user jobs. Finally, we present the performance advantages of the proposed wireless HPC system under more practical assumptions such as different cabinet capacities and diverse subtopology packings.

show abstract

“…Furthermore, realistic models of PBS 2 (or Torque 3 ) and OAR 4 are built-in. The quality of the results obtained during the validation of this work allows us to envisage using it as a performance prediction tool which could be embedded in the grid middleware to schedule its computing tasks, as described in [7].…”

Section: Introductionmentioning

confidence: 99%

Simbatch: An API for Simulating and Predicting the Performance of Parallel Resources Managed by Batch Systems

Caniou

2009

Euro-Par 2008 Workshops - Parallel Processing

View full text Add to dashboard Cite

The study of scheduling algorithms for parallel tasks in a grid computing context either neglects local reservation systems which manage parallel resources, either suppose that they use a First Come First Served strategy, or the experimental model does not handle parallel tasks. In this report, we describe an API built in the grid simulation tool Simgrid. It offers core functionalities to simulate in a realistic way parallel resources and batch reservation systems. Simbatch simulation experiments show an error rate inferior to 1% compared to real life experiments conducted with the OAR batch manager.

show abstract

Using Moldability to Improve the Performance of Supercomputer Jobs

Cited by 35 publications

References 27 publications

The cost of doing science on the cloud: The Montage example

The cost of doing science on the cloud: The Montage example

Job Mapping and Scheduling on Free-Space Optical Networks

Simbatch: An API for Simulating and Predicting the Performance of Parallel Resources Managed by Batch Systems

Contact Info

Product

Resources

About