How we did it

dAll throughput and interoperability tests, both homogeneous (client and router from the same company) and heterogeneous (all combinations of all clients and routers), were performed using the same configuration.

The basic test geometry consisted of the access point under test placed in a basement of our test house, with the client located on the second floor. This configuration involves a path that is essentially straight up, but the actual propagation mode(s) involved cannot be predicted nor easily analyzed. This geometry allowed us to test each configuration under simulated long-range conditions, with the signal nominally having to pass through two floors regardless. We completed three 180-second runs, averaging the results to get a final score for each pair of products.

We ran the Iperf client on a Dell Dimension 8400 desktop directly connected via Gigabit Ethernet to switch Port 1 of the access point under test. We ran the Iperf server on an IBM X40 notebook on the second floor. This computer was placed on a turntable revolving at approximately 1 RPM so as to factor out any anomalous antenna orientations, and to simulate, at least to a small degree, mobility. Both client and server PCs ran Windows XP Pro.

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

dAll throughput and interoperability tests, both homogeneous (client and router from the same company) and heterogeneous (all combinations of all clients and routers), were performed using the same configuration.

The basic test geometry consisted of the access point under test placed in a basement of our test house, with the client located on the second floor. This configuration involves a path that is essentially straight up, but the actual propagation mode(s) involved cannot be predicted nor easily analyzed. This geometry allowed us to test each configuration under simulated long-range conditions, with the signal nominally having to pass through two floors regardless. We completed three 180-second runs, averaging the results to get a final score for each pair of products.

We ran the Iperf client on a Dell Dimension 8400 desktop directly connected via Gigabit Ethernet to switch Port 1 of the access point under test. We ran the Iperf server on an IBM X40 notebook on the second floor. This computer was placed on a turntable revolving at approximately 1 RPM so as to factor out any anomalous antenna orientations, and to simulate, at least to a small degree, mobility. Both client and server PCs ran Windows XP Pro.

Both computers were configured by making sure all of the latest (as of Aug. 30) Microsoft Windows patches and updates were applied, power-saving was disabled so that the machines would always run at full speed, client and server IP addresses were statically assigned, and internal wireless adapters were disabled except for the first baseline test. All drivers on the client notebook were provided by the vendors and were uninstalled after each client adapter was tested. The client was also rebooted after each new installation.

Our second test involved exactly the same configuration and geometry, but used an Iperf stream simulating UDP video traffic. We thus tested both TCP, typical of most Web applications, and UDP, often used for video distribution, under the same set of conditions – only the benchmark parameters changed. We tested only homogenous configurations in the case, interoperability having been established by testing every combination of client and router using the first test.

The Iperf command lines used were as follows:
TCP:
(client) iperf –c 192.168.1.101 –w 128k –i .5 –r –t 90 ><file>.txt
(server) iperf –s –w 128k
UDP:
(client) iperf –c 192.168.1.101 –u -w 768k –i .5 –t 180 -b 100m ><file>.txt
(server) iperf –s –u –w 768k

For our qualitative video evaluation, we used the popular SlingBox Pro, a personal IP-TV system to watch actual video from a high-definition (HD) cable box. A corresponding SlingPlayer application was installed on the computer connecting to the SlingBox over the wireless link. We chose the SlingBox Pro because of its HD capabilities and fairly high (3Mbps) continuous bit rate. In this case, the SlingBox Pro and the access point were located on the second floor of the house, and testing involved walking away from the front of the house until we noticed the video occasionally freezing. We’d then carefully continue walking away until we could no longer see the video moving, and judged the last location where continuously-moving video was noted as the limit.