Development of Comprehensive Parallel I/O Benchmarking Tool for HPC Cluster File Systems

Abstract

High Performance Computing Clusters uses parallel file systems such as Lustre, GlusterFS for better I/O throughput with multiple clients. These file systems are implemented over high speed interconnects like 10G, Quadrics, Myinet and Infiniband. Parallel scientific applications require high performance I/O support from underlying file systems. The objective of this work is to evaluate and analyze performance of parallel file systems in HPC cluster with different configuration parameters with development of a comprehensive bench-marking tool. The developed tool measures the performance of different data I/O and meta-data operations under various workloads. This paper briefly describes various cluster file systems, parallel I/O approaches and different Message Passing Interface (MPI) implementations supporting MPI-IO and throughput results for different I/O scenarios.

Authors and Affiliations

M Swarnalatha , P. Durgadevi, T. Chindrella Priyadharshini, M. Hemalatha

Keywords

Related Articles

A Video Fusion Using Wavelet Transformation Techniques

Image processing techniques mainly focus upon enhancing the quality of an image or a set of images and to derive the maximum information from them. Image Fusion is such a technique of producing a superior quality image...

Green computing interchange Switching function

This research is about a method for enabling user of a computing system to measure and compare the green efficiency of a set of resources used for computing task. With the use of these information, the user can select a...

An Automatic Repelling System to Reduce Human Elephants Conflicts Using Sensors

In major areas, Human-Animal Conflict occurs due to the shortage of food and shelter. Due to the human population explosion and deforestation, forest areas were completely destroyed and this results in Human-Animal Conf...

Experimental Study on the Behavior of Copper Slag as Partial Replacement of Fine Aggregate in Concrete

In present arena, rapid urbanization has created a huge demand for natural sand hence made it even more expensive. Alternative materials in all forms of constructions are introduced to reduce the pressure on natural mat...

Structure Adaptable Neural Network Based Nonlinear Tracking Control of Kinematically Redundant Robot Manipulator

the performance of neural network based network in nonlinear tracking controller is like a lot of added absolute apple dynamical systems, adeptness systems are non-linear appropriately crave a acceptable adjustment of a...

Download PDF file
  • EP ID EP20359
  • DOI -
  • Views 254
  • Downloads 3

How To Cite

M Swarnalatha, P. Durgadevi, T. Chindrella Priyadharshini, M. Hemalatha (2015). Development of Comprehensive Parallel I/O Benchmarking Tool for HPC Cluster File Systems. International Journal for Research in Applied Science and Engineering Technology (IJRASET), 3(5), -. https://europub.co.uk./articles/-A-20359