ABSTRACT
Contents 17.1 Introduction ................................................................................................................... 438 17.2 Basic Terminologies and Background ............................................................................. 439
17.2.1 MPI and OpenMP ............................................................................................. 439 17.2.2 Performance Measurement with HPL ................................................................. 440
17.3 Architecture of Cluster ................................................................................................... 440 17.3.1 Machine Specifications ....................................................................................... 440 17.3.2 Configuration of Cluster ..................................................................................... 442
17.4 Research Methodology and Details of Our Experiment ................................................. 442 17.4.1 Single Node Test ................................................................................................. 443 17.4.2 Overall Cluster Performance ............................................................................... 443
17.5 Observations and Analysis ............................................................................................. 446 17.6 Concluding Remarks and Future Research Direction .................................................... 449 References ............................................................................................................................... 449
17.1 Introduction High-performance clusters can be considered as part of the underlying components for nextgeneration networks (NGNs). As cluster size and core numbers continue to expand, NGN infrastructure could heavily use such systems. It can be envisioned that the NGNs will generate huge amounts of data that may need to be processed by high-performance computing clusters. erefore, an understanding of clusters and the different programming paradigms (that are available) is needed for optimal utilization. is chapter is written from the philosophy that the gained practical knowledge on an implemented cluster would help development of future technologies and mechanisms to deal with related issues.
