New WLAN on Campus: Page 17 of 31

For throughput testing, we used NetIQ's Chariot 4.3 network-performance application and Dell Latitude laptops with 256 MB of RAM running Windows 2000. We used Chariot's TCP-based unidirectional long file receive (filercvl) test script with 100 iterations of a 1-MB file. We tested both 802.11b and 802.11a capabilities with one and then four clients of each standard.

We ran each test in two directions: from the wired endpoint to the wireless client or clients (downstream); and from the wireless endpoint or endpoints to the wired endpoint (upstream). Laptops were located about five feet from the APs (access points).

We gave the vendors a choice of which client card to use for single-station testing, but confined them to a set array of cards for our four-station tests. Clients for our four-station 11b testing consisted of a Symbol Spectrum24 card, a Netgear 11a/b card and two Cisco 350 cards. For 802.11a four-station testing, we used a Proxim Harmony 802.11a card, an Intel card, a Netgear 11a/b card and a Proxim 11a/b/g ComboCard. We used a variety of client cards in our four-station throughput tests to reflect a real-world scenario and to assess basic interoperability.

Note that many factors influence the performance of real-world wireless networks. Our performance measurements provide only a crude comparison between products operating in a best-case lab environment. Results for all performance tests are shown on here.

WLAN Features click to enlarge

We performed range testing for the APs in a central, ceiling-level location in a large walled classroom/office building constructed in the mid-1980s. Most walls were Sheetrock over metal studs, and most doors were metal with some glass. We restricted APs to minimum data rates of 5.5 Mbps for 11b and 12 Mbps for 11a, then measured maximum ranges based on packet loss. Comparative product ranges are illustrated left.