How we tested the WLAN gear

We assessed enterprise-grade 802.11n systems in terms of pure 802.11n throughput and latency; mixed-mode (802.11n plus both 802.11g and 802.11a clients) throughput and latency; "WiMix" traffic handling, which comprises a combination of common enterprise application types; power over Ethernet consumption; and features. A complete, more detailed version of this methodology is available here.

We asked participating vendors to supply eight dual-radio (2.4- and 5-GHz) access points and placed each access point inside its own shielded RF chamber. We attached the VeriWave WaveTest WT-90 test instrument to the chambers (and within the chambers, to the antenna connectors of each access point) using multiple cables, one for each spatial stream. This setup ensured repeatability and freedom from outside interference sources. The VeriWave WT-90 also offered gigabit Ethernet traffic from up to four ports.

To measure 802.11n throughput, we defined a single SSID on the 5GHz radio for each of eight access points. We also configured access points and tested instruments to use 40MHz of bandwidth and short (400-nanosec) guard intervals to obtain the highest possible rate.

We offered test traffic at varying loads to determine the throughput rate (the maximum rate at which the access points dropped none of the offered frames). We repeated these tests nine times: Three times each offering downstream, upstream and bidirectional traffic, using three different frame sizes, using 88-, 512- and 1,518-byte frames. We also measured average and maximum latency and jitter at the throughput rate. The test duration in all cases was 60 seconds.

In the mixed-mode (or "convergence") tests, the goal was to determine throughput and latency for an access point concurrently handling a mix of 802.11n, 802.11g and 802.11a clients. We enabled both 2.4- and 5-GHz radios on four access points for these tests and associated 16 802.11n clients and four 802.11a or 802.11g clients to each radio. We used standard (800-nanosec) guard intervals on the 2.4-GHz radios and short (400-nanosec) guard intervals on the 5-GHz radios.

As in the pure 802.11n tests, we used a binary search algorithm to determine the system's throughput rate. Once again we ran separate tests for downstream, upstream and bidirectional traffic, and again used the same three frame lengths. We also measured average and maximum latency and jitter.

The WiMix tests modeled a mixture of four common enterprise applications: Web traffic, Windows file traffic (using the SMB/CIFS protocols), Windows media traffic and VoIP.

We defined a separate service-level agreement (SLA) for each traffic class. VoIP traffic had to exhibit an R-value of at least 78, which is generally considered the threshold for toll-quality calls. The access points had to forward Windows Media traffic with latency and jitter of no more than 500 millisec, while also dropping no more than 2% of packets. And to avoid "starving" the Web and Windows traffic classes, the access points had to forward each class at a rate equivalent to at least 5% of the intended load.

The goal of the WiMix test was to determine the highest traffic load at which the access points enforced the SLAs for all four traffic classes. As in the convergence tests, we ran the WiMix tests on four access points with both the 2.4- and 5-GHz radios active, and the access points fielded traffic from a mix of 802.11n and non-802.11n clients.

We measured power over Ethernet consumption on a single access point. We enabled both radios on the access point and asked vendors to set each radio to draw maximum power.

Using a Fluke 87V multimeter, a Fluke i30 DC current clamp and a specially constructed Ethernet cable, we then measured voltage from the power source, as well as current drawn (in milliamps) by the access point. We first measured current while the access point was idle. Then we used the VeriWave WT-90 to run maximum forwarding rate tests with both 88- and 1,518-byte frames and measured maximum power draw during each test. We also recorded the maximum forwarding rate as well as the power usage for each frame size.

To assess the final area, system features, we asked vendors to complete a detailed questionnaire covering system capabilities and pricing. We did not verify every answer to this questionnaire.

< Return to main test: 802.11n gear 10 times faster than current Wi-Fi offerings >

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Must read: 10 new UI features coming to Windows 10