Chair: Pak Lui
HPC clusters have become the preferred solution for high-performance systems as they provide an efficient performance compute solution based on industry standard hardware. The main benefits of clusters are affordability, flexibility and availability. A cluster uses the aggregated power of compute server nodes to form a high-performance solution for parallel applications. When more compute power is needed, it can be simply achieved by adding more server nodes to the cluster. The way HPC clusters are architected (i.e. CPUs, GPUs, high-speed interconnects, memory, compilers, libraries, storage) has a great influence on the overall application performance and productivity. By providing one of the largest database of applications best practices and guidelines, the HPC|Works subgroup enables HPC users to review all the capabilities for building balances and scalable HPC systems, and tune their systems for maximum performance. The applications database can be viewed here.
The following provides guidelines for application installations on HPC clusters, including the required software components, how to build and how to run simple benchmarks. The actual results of the benchmarks, profiling information and more are posted on the best practices page as listed above.