How we did it
Cisco's Aironet access point knocks us out in the 802.11b arena, but others aren't far behind
|
|
|||
|
|
It's hard to look at a wireless LAN access point by itself. It's a part of a network, and as part of an enterprisewide network it extends communications beyond the LAN. We were concerned with performance, security, manageability, interoperability, stability and scalability.
For performance testing, we connected each access point to our 100Base-T network and configured it using the vendor's default setup. Then, wherever possible, we installed the same vendor's wireless NICs in our eight test machines. Because several vendors did not send us eight NICs, we substituted Orinoco NICs for the missing NICs. In all cases, we could test with at least four of the vendors' NICs.
We ran throughput tests on one, two, four and eight machines simultaneously. To test IPX performance, we used Novell's venerable Perform3 diagnostic, a part of its LANTest suite. To test IP performance, we used scripted FTP transfers using RhinoSoft's FTP Voyager to connect to a local FreeBSD Unix host. In all cases, there were no appreciable differences between the four- and eight-node tests, which tells us that three or four nodes can saturate an 802.11b channel.
Next we checked range. We put a Symbol NIC with an external antenna into a PC in the next building and checked throughput to that machine with each access point. There were about 60 feet, and four walls, separating the access point and the PC.
For interoperability testing, we used an IBM ThinkPad T20 and installed a different NIC in each PC. Then we swapped the access points and made sure all the NICs could connect to each access point. We ran an abbreviated performance test and tabulated the results. This also gave us a good idea of how much performance came from the NICs and how much from the access points.
We also evaluated ease of use, setup, security and stability of the products. In several cases, we found that products we were evaluating were made by the same manufacturer, but the differences in drivers made big differences in the results we observed. We evaluated security by looking at options available to help the system manager secure the LAN, and what the implications of these tools were on manageability. We gave the products points for each of the security features they had - ESSID, Access Lists, RADIUS support, single-point management, 40-bit WEP, 128-bit WEP and central WEP key management.
We gave points to the management category for features including remote power, single point management of nodes, centralized WEP key management, and each way of accessing an access point to control and manage it including serial cable, telnet, proprietary consoles, Web interfaces and FTP.
RELATED LINKS
Avery is the founder of Gunnison Territory Network Consultants, a small firm specializing in network design, management and administration. He can be reached at mavery@mail.otherwhen.com.
