Large clusters composed of high-end commodity components now offers matching or superior performance for many parallel codes relative to proprietary HPC systems for a fraction of the cost. Individual researchers are demanding, and in some cases building, small local clusters because they offer the best price/performance solution for many scalable codes. However, for users to take full advantage of clusters as computational resources, they need to fully understand the issues involved in: developing codes for clusters, whether they have Intel Xeon/Pentium4, Intel Itanium, or AMD Opteron processors; optimizing codes for maximum performance on nodes with commodity processors and memory; and porting codes between different clusters, and between clusters and traditional 64-bit HPC platforms, while retaining high efficiency and scalability. This project is exploring all of the issues and producing technical documentation to assist users with developing and executing applications on clusters.