For Web performance testing, we used RadView Software's WebLoad 4.5 to see how many transactions per second a server could handle. We selected a 24-KB file with three 8-KB elements, which we have used in other tests (click here for our review of WebLoad 4.5).

WebLoad 4.5, RadView Software, www.radview.com.

Hard-Drive Subsystem Performance Click here to enlarge

To get good system hard-disk subsystem performance benchmarks, we tested with Intel's IOmeter version 1998.10.08. We set up the tests to run with the following parameters:

For the maximum throughput test, we changed the transfer-request size to 64 KB, the percent read/write distribution to 100 percent read and the percent random/sequential distribution to 100 percent sequential.

For the maximum I/O rate test, we changed the transfer-request size to 512 bytes, the percent read/write distribution to 100 percent read and the percent random/sequential distribution to 100 percent sequential.

(Intel does not endorse IOmeter results.)

Intel Corp., IOmeter version 1998.10.08, www.intel.com.

For validation of network-interface performance, we went with NetIQ Corp.'s Chariot 3.2. We used the pregenerated script filesndl (FileSendLong). Employing six client machines, we sent data to the server for the receive test, sent data to the client machines for the send test and used an even mix for the bidirectional tests. Data was collected on a separate machine running Microsoft Windows NT 4.0 and the NetIQ Chariot console.

NetIQ Corp., Chariot 3.2, www.netiq.com.