Network Performance: Considering The User

10 Pages 2465 Words

nce over the network when the measurements fail to offer a solution.

COMPONENT AND DEVICE TESTING


Discussion: Most of the test equipment used for benchmarking, although quite rich in features, are basically stress testers that can send constant rate data through a device, then record valuable measurements. For example, a valid test method involves a test set transmitting frames to an input port on the device under test (DUT) and then measuring those frames on an output port (with the same test set, of course). Overhead traffic like keep-alive frames and routing updates are omitted from the receive count, ensuring a valid measurement of frame loss and errors. When multiple devices are to receive test traffic, as is frequently the case with interoperability and design validation testing, the DUT is made to route the test frames, as is the case in the real world. Sophisticated test equipment can accomplish this by sending test data to and from multiple protocol addresses.

On the particular subject of benchmarking, certain standards and commonly accepted practices are critical to the quality of the results that are offered to network designers. Test trials, for example, should be conducted using a representative range of frame sizes; a good set would include frames of the minimum and maximum transmission unit for the associated topology, with a healthy sampling of sizes in between.

Test duration is another item for serious consideration, particularly when benchmarking a new device. The following text, which refers to the duration of test trials (each test consists of multiple trials), is from RFC 2544, Benchmarking Methodologies for Network Interconnect Devices, Bradner/McQuaid (1999):

“The aim of these tests is to determine the rate continuously supportable by the DUT. The actual duration of the test trials must be a compromise between this aim and the duration of the benchmarking test suite. The duration of...