AIM Technology bases its Server Benchmark for Windows NT on a mix-and-load modeling scenario, in which you run a predefined primary application mix that resembles the tasks an NT server most commonly performs. You record the server's performance as you incrementally increase the load of the application mix to determine peak and sustained performance metrics for the system.
The AIM WNT Peak Performance metric measures the highest attainable performance a system can achieve. The system must run the CPU, RAM, and disk caching at maximum to attain peak performance. The AIM WNT Sustained Performance metric measures the maximum system capacity at the point when additional tasks significantly hinder performance. The difference between the AIM WNT Peak Performance metric and the AIM WNT Sustained Performance metric is similar to the difference between a drag race and a rally covering hundreds of miles. Servers must perform continuously and consistently. Therefore, the AIM WNT Sustained Performance metric is more relevant than the AIM WNT Peak Performance metric in determining a server's performance in a real-world network installation.
AIM has two server benchmarks for NT: the Domain Server Mix and the File Server Mix. The Domain Server Mix includes 54 tests that simulate the duties a typical domain server performs (e.g., light file transfers; network routing; packet forwarding; email; and shared applications, such as spreadsheets, word processing, and network maintenance). The File Server Mix includes 38 tests that simulate the duties a typical gateway file server performs (e.g., name, mail, and network services). For the 8-way server tests, the Windows NT Magazine Lab ran the Domain Server Mix.
—Brian Gallagher