That would depend on the number of elements and components you plan on monitoring, but I would say best practice would be a physical SQL server with RAID array and a healthy amount of memory. If you were monitoring 100,000 SAM components and NPM elements, I would recommend that array be at least a six or eight spindle SAS RAID 10 with the TempDB located on SSD and your transaction logs broken out onto it's own mirror, RAID-0, or RAID 10. You would probably want at least 8-16 CPU Cores in that SQL server and 64-128GB of RAM. Smaller environments can most certainly get away with far less hardware, and even larger environments could benefit from more spindles to the log, tempDB, and data volumes, moving entirely to SSD, or adding additional CPU cores and memory.
↧