Myrinet networks: a performance study

Qian, Yiming; Afsahi, Ahmad; Zamani, Reza

doi:10.1109/nca.2004.1347794

Cited by 5 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This implies that all messages are transmitted over the interconnection network at the same transmission rate having same transmission links during each transmission between two tasks and the links between any two processors change with each communication edge in order to avoid contention delay. Although, Gunaratne et al and Qian et al have shown that the energy consumed by high-speed interconnects in either idle or fully utilize mode is almost the same (Gunaratne et al, 2008;Qian et al, 2004). In the present work, we have considered interconnection to be dynamic.…”

Section: Energy Consumption Limits (E G )mentioning

confidence: 98%

See 1 more Smart Citation

An Analytical Model for Resource Characterization and Parameter Estimation for DAG-Based Jobs for Homogeneous Systems

Sajid

Raza

2015

International Journal of Distributed Systems and Technologies

View full text Add to dashboard Cite

High Performance Computing (HPC) systems demand and consume a significant amount of resources (e.g. server, storage, electrical energy) resulting in high operational costs, reduced reliability, and sometimes leading to waste of scarce natural resources. On one hand, the most important issue for these systems is achieving high performance, while on the other hand, the rapidly increasing resource costs appeal to effectively predict the resource requirements to ensure efficient services in the most optimized manner. The resource requirement prediction for a job thus becomes important for both the service providers as well as the consumers for ensuring resource management and to negotiate Service Level Agreements (SLAs), respectively, in order to help make better job allocation decisions. Moreover, the resource requirement prediction can even lead to improved scheduling performance while reducing the resource waste. This work presents an analytical model estimating the required resources for the modular job execution. The analysis identifies the number of processors required and the maximum and minimum bounds on the turnaround time and energy consumed. Simulation study reveals that the scheduling algorithms integrated with the proposed analytical model helps in improving the average throughput and the average energy consumption of the system. As the work predicts the resource requirements, it can even play an important role in Service-Oriented Architectures (SOA) like Cloud computing or Grid computing.

show abstract

Section: Energy Consumption Limits (E G )mentioning

confidence: 98%

“…As an example, Myrinet-2000 does not have any power management mechanism. The energy consumed by Myrinet-2000 network switches remains unchanged regardless of the network traffics (Qian et al, 2004). The dynamic interconnects are likely to appear in the near future in order to save energy in the communication networks.…”

Section: Energy Consumption Limits (E G )mentioning

confidence: 99%

An Analytical Model for Resource Characterization and Parameter Estimation for DAG-Based Jobs for Homogeneous Systems

Sajid

Raza

2015

International Journal of Distributed Systems and Technologies

View full text Add to dashboard Cite

show abstract

Towards A Model-Based Autonomic Reliability Framework for Computing Clusters

Dubey¹,

Nordstrom²,

Keskinpala³

et al. 2008

Fifth IEEE Workshop on Engineering of Autonomic and Autonomous Systems (Ease 2008)

View full text Add to dashboard Cite

One of the primary problems with computing clusters is to ensure that they maintain a reliable working state most of the time to justify economics of operation. In this paper, we introduce a model-based hierarchical reliability framework that enables periodic monitoring of vital health parameters across the cluster and provides for autonomic fault mitigation. We also discuss some of the challenges faced by autonomic reliability frameworks in cluster environments such as non-determinism in task scheduling in standard operating systems such as Linux and need for synchronized execution of monitoring sensors across the cluster. Additionally, we present a solution to these problems in the context of our framework, which utilizes a feedback controller based approach to compensate for the scheduling jitter in non realtime operating systems. Finally, we present experimental data that illustrates the effectiveness of our approach.

show abstract