Computational Technology Resources - CSETS

Keywords: high performance computing, architectures, software, solutions.

Summary

High performance computing is currently undergoing an interesting phase of change. In terms of architectures for high performance computing we see both convergence and divergence. The economic crisis of recent years has certainly had an impact on the community. At the same time we see an interesting political situation in which more and new players are aiming at reaching an international level in high performance computing. The concept of hybrid computing has somewhat lost its appeal during recent years. At the same time accelerator technology is being used to increase performance in the TOP500 list. The impact on computational engineering is dramatic. In terms of software the main challenge is to keep pace with the hardware development.

We see a number of national activities which are targeting an increase in computer performance at the national level. This political support has led to a number of big national projects in China, the European Union, Japan and the US. One of the big targets for all of these projects is the Exascale System. This is expected to be available in 2018.

The main challenge in using such a system will be both in the number of cores that such a system will be built from and the hierarchical concept in which these cores will be used. Hence it will be important to find the right programming methods for such systems. Specifically for engineering applications the challenge will be huge.

The basic rules for successfully using a modern supercomputer are simple and difficult to obey at the same time:

Avoid communication. By reducing communication at all we reduce the overhead for exchanging information in a loosely coupled hardware system
Communicate as locally as possible. If we can limit most of our communication to the processor or the node we can again avoid using a loosely coupled system.

In essence we need to revisit some of our basic concepts for simulation in engineering. Experience shows that moving from traditional mesh based concepts (such as finite element or finite volume) to Lattice-Boltzmann methods can help to improve the performance of flow simulations substantially. We have to find more of these concepts that fit much better into the upcoming architectures.

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £95 +P&P)

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Computational Science, Engineering & Technology Series ISSN 1759-3158 CSETS: 27 TRENDS IN PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING Edited by: P. Iványi, B.H.V. Topping Chapter 8 High Performance Computer Simulation for Engineering: A Review M.M. Resch High Performance Computing Center Stuttgart (HLRS), University of Stuttgart, Germany doi:10.4203/csets.27.8 purchase the full-text of this chapter Full Bibliographic Reference for this chapter M.M. Resch, "High Performance Computer Simulation for Engineering: A Review", in P. Iványi, B.H.V. Topping, (Editors), "Trends in Parallel, Distributed, Grid and Cloud Computing for Engineering", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 8, pp 177-186, 2011. doi:10.4203/csets.27.8 Keywords: high performance computing, architectures, software, solutions. Summary High performance computing is currently undergoing an interesting phase of change. In terms of architectures for high performance computing we see both convergence and divergence. The economic crisis of recent years has certainly had an impact on the community. At the same time we see an interesting political situation in which more and new players are aiming at reaching an international level in high performance computing. The concept of hybrid computing has somewhat lost its appeal during recent years. At the same time accelerator technology is being used to increase performance in the TOP500 list. The impact on computational engineering is dramatic. In terms of software the main challenge is to keep pace with the hardware development. We see a number of national activities which are targeting an increase in computer performance at the national level. This political support has led to a number of big national projects in China, the European Union, Japan and the US. One of the big targets for all of these projects is the Exascale System. This is expected to be available in 2018. The main challenge in using such a system will be both in the number of cores that such a system will be built from and the hierarchical concept in which these cores will be used. Hence it will be important to find the right programming methods for such systems. Specifically for engineering applications the challenge will be huge. The basic rules for successfully using a modern supercomputer are simple and difficult to obey at the same time: Avoid communication. By reducing communication at all we reduce the overhead for exchanging information in a loosely coupled hardware system Communicate as locally as possible. If we can limit most of our communication to the processor or the node we can again avoid using a loosely coupled system. In essence we need to revisit some of our basic concepts for simulation in engineering. Experience shows that moving from traditional mesh based concepts (such as finite element or finite volume) to Lattice-Boltzmann methods can help to improve the performance of flow simulations substantially. We have to find more of these concepts that fit much better into the upcoming architectures. purchase the full-text of this chapter (price £20) go to the previous chapter go to the next chapter return to the table of contents return to the book description purchase this book (price £95 +P&P)
Back to top	©Civil-Comp Limited 2023 - terms & conditions