Predicting the performance impact of different fat-tree configurations

Jain, Nikhil; Bhatelé, Abhinav; Howell, Louis H.; Böhme, David; Karlin, Ian; León, Edgar A.; Mubarak, Misbah; Wolfe, Noah; Gamblin, Todd; Leininger, Matthew L.

doi:10.1145/3126908.3126967

Cited by 38 publications

(13 citation statements)

References 27 publications

(39 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…TraceR performs packet-level simulation of MPI applications on HPC networks. It efficiently and accurately predicts the performance of important applications on networks with different properties as previous work [13] has extensively validated.…”

Section: Methodsmentioning

confidence: 96%

Mitigating Inter-Job Interference via Process-Level Quality-of-Service

Savoie

Lowenthal

Supinski

et al. 2021

ACM Trans. Parallel Comput.

Self Cite

View full text Add to dashboard Cite

Jobs on most high-performance computing (HPC) systems share the network with other concurrently executing jobs. Network sharing leads to contention that can severely degrade performance. This article investigates the use of Quality of Service (QoS) mechanisms to reduce the negative impacts of network contention. QoS allows users to manage resource sharing between network flows and to provide bandwidth guarantees to specific flows. Our results show that careful use of QoS reduces the impact of network contention for specific jobs, resulting in up to a 40% performance improvement. In some cases, it completely eliminates the impact of contention. It achieves these improvements with limited negative impact to other jobs; any job that experiences performance loss typically degrades less than 5%, and often much less. Our approach can help ensure that HPC machines maintain high levels of throughput as per-node compute power continues to increase faster than network bandwidth.

show abstract

Section: Methodsmentioning

confidence: 96%

Mitigating Inter-Job Interference via Process-Level Quality-of-Service

Savoie

Lowenthal

Supinski

et al. 2021

ACM Trans. Parallel Comput.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Many modern HPC interconnects are based on either fat-tree or dragonfly topologies, or variants [11,12]. These topologies admit a number of variations, such as multi-rail or multi-plane fat trees [13], and different link widths or speeds. The design of the network affects almost every aspect of its performance, especially its sensitivity to congestion.…”

Section: Contentsmentioning

confidence: 99%

Understanding congestion in high performance interconnection networks using sampling

Taffet

Mellor-Crummey

2019

Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

Understanding Congestion in High Performance Interconnection Networks Using Sampling by Philip A. Taffet The computational needs of many applications outstrip the capabilities of a single compute node. Communication is necessary to employ multiple nodes, but slow communication often limits application performance on multiple nodes. To improve communication performance, developers need tools that enable them to understand how their application's communication patterns interact with the network, especially when those interactions result in congestion. Since communication performanceis difficult to reason about analytically and simulation is costly, measurement-based approaches are needed. This thesis describes a new sampling-based technique to collect information about the path a packet takes and congestion it encounters. Experiments with simulations show that this strategy can distinguish problems with an application's communication patterns, its mapping onto a parallel system, and outside interference. We describe a variant of this scheme that requires only 5-6 bits of information in a monitored packet, making it practical for use in next-generation networks.

show abstract

“…When a computational task arrives at a physical machine, it may require a variety of meteorological data at different locations for computation. Therefore, this paper uses fattree topology structure to construct the meteorological database of MCP [20].…”

Section: Meteorological Cloud Platform Over Big Datamentioning

confidence: 99%

A big data placement method using NSGA-III in meteorological cloud platform

Ruan

Huang³

et al. 2019

J Wireless Com Network

View full text Add to dashboard Cite

Meteorological cloud platforms (MCP) are gradually replacing the traditional meteorological information systems to provide information analysis services such as weather forecasting, disaster warning, and scientific research. However, the explosive growth of meteorological data resources has brought new challenges to the placement and management of big data in MCP. On the one hand, managers of MCP need to save energy to achieve cost savings. On the other hand, users need shorter data access time to improve user's experience. Hence, a big data placement method in MCP is proposed in this paper to deal with challenges above. First, the resource utilization, the data access time, and the energy consumption in MCP with the fat-tree topology are analyzed. Then, a corresponding data placement method, using the improved non-dominated sorting genetic algorithm III (NSGA-III), is designed to optimize the resource usage, energy saving, and efficient data access. Finally, extensive experimental evaluations validate the efficiency and effectiveness of our proposed method.

show abstract

Predicting the performance impact of different fat-tree configurations

Cited by 38 publications

References 27 publications

Mitigating Inter-Job Interference via Process-Level Quality-of-Service

Mitigating Inter-Job Interference via Process-Level Quality-of-Service

Understanding congestion in high performance interconnection networks using sampling

A big data placement method using NSGA-III in meteorological cloud platform

Contact Info

Product

Resources

About