Air Time: Understanding 802.11 Interference
Under conditions of heavy network utilization and co-channel interference, today's wireless LAN products behave unpredictably. Over time, these problems will become more critical.
May 24, 2007
Last November, Network Computing ran a feature story examining the performance of Wi-Fi products from Meru Networks and Cisco Systems. Their offerings differed fundamentally in how they manage contention. Meru markets its unique contention-management algorithms as its "Air Traffic Control" system. On most Wi-Fi networks, APs and wireless clients have equal rights to transmit. On Meru networks, APs and controllers exert more control, prioritizing traffic in ways it says offer significant advantages, particularly for converged voice and data environments.
Our initial testing confirmed Meru's value proposition. Meru bested Cisco in fairly managing contention between Vo-Fi phones and notebook computers. Cisco cried foul, asserting that Meru was playing fast and loose with standards, manipulating fields associated with 802.11's virtual carrier sense system. Packet traces confirmed that certain duration field values in Meru packets were out of spec. It was far short of proof, but it looked like a smoking gun.
Packet traces aside, we were more interested in real-world impact. We rigged up a test bed and threw a mix of 802.11b/g traffic at Cisco and Meru APs running on the same RF channel. Meru's performance dropped by 50 percent, as one might expect, but Cisco's cratered. Cisco pointed to this as proof that Meru was cheating. Meru blamed it on Cisco's poor software engineering.
We published what we knew at the time, reporting the favorable Meru test results as well as the coexistence results and Cisco's allegations of standards noncompliance. And we gave both vendors the opportunity to offer a formal response (see Cisco Vs. Meru: The Vendors Speak).
Cisco then brought its allegations to the Wi-Fi Alliance late last year, asserting that Meru was violating the alliance's good-neighbor policy. This was not the first time Meru had been challenged, but the result was the same: The alliance refused to rescind Meru's certification. Cisco to this day refuses to back down from its allegations.Meanwhile, back at the lab, a review of Aruba Networks' latest offerings provided us with the opportunity to add another vendor to our good-neighbor test bed. When we tested with Aruba and Cisco APs installed on the same channel, the results were similar to what we encountered with Meru. Although Cisco is still investigating our results, a product manager suggested that Aruba's unique use of the 802.11 PCF timing algorithm might be the cause. Aruba denies using PCF timing and claims its implementation is simply more efficient than Cisco's.
The story doesn't end there. In our initial testing, we used notebooks with Atheros-based NICs. This time, we ran additional tests using NICs based on Broadcom and Intel chipsets. The results were astonishing. Not only did we find significant variation in the each AP's performance, we also found significant differences in performance using various client chipsets. While Meru performed quite well with Atheros and Broadcom clients, for instance, it performed poorly with Intel clients.
The implications of these tests are alarming. Under conditions of heavy network utilization and co-channel interference, today's wireless LAN products behave unpredictably. Given the relatively low utilization and user density of enterprise Wi-Fi networks, it's not surprising that these problems don't keep network administrators up at night. But over time, as wireless LANs assume a more central role and their density increases, these problems will become more critical.
Our test results cast a shadow over Cisco's leadership position in the enterprise WLAN market. For now, we'll give Cisco the benefit of the doubt as we work with company reps to understand why its equipment performs so poorly when running side by side with competitors' gear. But it's clear that organizations like the Wi-Fi Alliance must bolster their testing to include more than base-level interoperability. If all these products are Wi-Fi certified, you have to wonder what that really means.
Dave Molta is a Network Computing senior technology editor. He is also assistant dean for technology at the School of Information Studies and director of the Center for Emerging Network Technologies at Syracuse University. Write to him at [email protected]1022
You May Also Like