Now that you have described the software environment, what about the hardware? Are you looking for large SMPs or a cluster of thin nodes? What are your availability and operational requirements? For the cluster, what kind of interconnect performance is needed? What is your expected data throughput for this application? You probably have decided these points already, but, if you are not set on them, the benchmark can influence your decision. You may even solicit advice or 'best' solution from the vendors.
Are there any extreme memory requirements, say, for the performance of the SGA (shared global area) or for allocating scratch space as tmpfs (a memory-based file system)?
For many Commercial benchmarks, the disk configuration is more important for performance than the memory. How much disk space is required for the database? Will you need an online backup of the database to ease the restoration of the database in between runs? What are the requirements for a logging file system? For HPTC, how much space does your out-of-core solver need? What is the configuration of the disks: RAID 5, SAN, just a bunch of disks (JBOD)? Do you have a corporate standard for file systems like UFS or VxFS? If so, is that required in the benchmark? Then state that.