The launch of the Grid Innovation Zone@NUS (GIZ@NUS) on the 18 March 2004 signified the beginning of a new chapter in Grid computing development at NUS. The collaboration between NUS, IBM, Intel and IDA brings together resources and expertise to promote and support the research, development, implementation and usage of Grid technology. The first article on GIZ@NUS provides more details on how NUS will benefit from this collaboration. Further details can also be found in the report by CNET.

With the advent of cluster computing technology and the availability of low cost cluster solution, more research computing applications are being deployed in a cluster environment rather than on a single shared-memory system. High-performance cluster computing is more than just having a large number of computers connected with high-bandwidth low-latency interconnect. To achieve the intended speed-up and performance, the application itself has to be well parallelised for the distributed-memory environment. In research computing, the MPI based Message Passing technique is the most popular technique used for parallel application development today. The article on Using MPI on Atlas provides some tips on how to write and run a parallel code using MPI. The third article on Performance of MPI-based Applications describes two user applications that have succeeded in achieving good performance on the Linux cluster provided by SVU. In the fourth article, one of our users will share his experience in the Simulation of Passive Fluid Mixing in Microchannels. Last but not least, researchers who are developing their own computational codes may wish to read about the advanced visualisation tools currently available at SVU. These tools are described in the last article.

Happy reading!

The objective of the collaboration among NUS, IBM, Intel and IDA is to set up a centre of expertise and resources to promote and support the use of Grid technology on campus. The Centre will be equipped with a 32-CPU Itanium2 based cluster, storage resources and various Grid computing tools. Users or research collaborators at the Centre will also have access to Grid computing related expertise provided by IBM. It is a centre where faculty members and students can go to be trained, test applications and leverage on the experience of the centre's resources. New ideas or solutions can be tested at the Centre before they are deployed in a production environment. Potential projects include development and testing of applications that will be deployed in the NUS Grid or the National Grid environment. Faculty members and students who are interested to explore the Grid technology for their research work are welcome to contact us at .

Message Passing is a popular technique used for writing parallel code. Various libraries exist and one of the most widely adopted is MPI (Message Passing Interface). In this article, we explore MPI using fortran language bindings. Read on if you want to learn how to write and run parallel code using MPI in general and on the atlas cluster in particular.

Many SVU users may not be aware that they can now run parallel C or FORTRAN programs on our latest Linux cluster with 16 compute nodes, each with two 3.06GHz Intel Xeon processors with Hyper Threading and high-speed Myrinet interconnects. Highly efficient compilers from Intel and Portland Group as well as MPI libraries are available for both C and FORTRAN to help users parallelise their codes.

Since the release of the cluster, SVU has helped a few users to parallelise and run their programs using the above tools and libraries. Lessons learned from these user programs show that parallel processing will speed up programs, provided that communications overheads are low(Read details). With our Linux cluster, programmers need only knowledge of MPI to create parallel programs, making it easier for anyone to program in parallel. For more information on how to access the cluster and to write parallel programs, please contact me at .

Fluid mixing at the microscopic scale has attracted great interest among researchers with the continuous development of microfluidic systems, such as lab-on-a-chip for biochemistry analysis, micro-reactors for production of organic compounds, etc. At the microscopic scale, fluid streams prefer to adopt laminar flow patterns, thus fluid mixing mainly depends on molecular diffusion which is very slow. This study aims to simulate fluid mixing at the microscopic scale using CFD method. The simulation results can be used as a prediction of the mixing performance and will aid to develop highly efficient micro-mixers. Read details...

If you are writing your own Computational Fluid Dynamics or Structural Mechanics codes to solve large and complicated research problem, you probably like to have a ready-made, off-the-shelve visualisation tools that will help you in analysing the simulation results. EnSight, a post-processing software that is general enough to be used for many computational science and engineering applications and yet sophisticated enough to be used as an advanced visualisation tool, is the right tool for you. The visualisation tool, which is also used by Scientists at NASA, provides some advanced scientific visualisation features such as particle plane clips, isosurfaces, load, transient data and particle trace animation. Some application examples can be viewed at http://www.ceintl.com/solutions/cfd.html.

EnSight can be accessed using the graphics workstations at the SVU (Supercomputing & Visualisation Unit) Visualisation Laboratory, 1^st level, Computer Centre. It can also be accessed using the remote visualisation system introduced recently. Through the remote visualisation system, you can access the powerful 64-bits graphics server right from your desktop or wireless laptop. The procedure on accessing the remote visualisation system is available at http://e-svu.nus.edu.sg/janus/Instruction.htm. We welcome you to send your feedback or query to .