Consumer Wi-Fi networks do not play well together when competing for the same Radio Frequency space. Learn how Standford researchers get them to cooperate.
Wi-Fi networks are everywhere, in particular apartment complexes or college dorms, where more often than not every resident runs a personal Wi-Fi network. That Radio Frequency (RF) density has a huge impact on how well each Wi-Fi connection (network Wi-Fi device to Wi-Fi client) performs. Interference and congested channels often lead to severely degraded performance.
That's not good when people expect their systems to perform well; HD movies should remain high definition and not pixilated; Skype calls should resemble normal phone conversations, and not sound like the other person is on the moon.
There are books that cover Wi-Fi interference and the problems it causes (I also wrote about it on TechRepublic). The quick version is that Wi-Fi uses a technology called Spread Spectrum. That means even though there are 14 channels on the popular 2.4 GHz frequency band, only three channels (1, 6, and 11) are truly non-interfering (Figure A). Imagine trying to coordinate all the Wi-Fi routers and Access Points (APs) in an apartment to prevent any channel interference.
There are solutions
There are commercially available systems that will optimize frequency usage preventing interference, but they are expensive -- approximately $500 - 1,000 for each device. To be fair, most of the newer consumer-grade Wi-Fi routers and APs have the ability optimize channel selection, but as Figure A shows, unused does not always mean noninterfering.
There is another solution, described in the paper SDN for Dense Home Networks (PDF), proposed by a team of Stanford University researchers who say they have figured out how to turn Wi-Fi crowding into an advantage: having lots of Wi-Fi routers or APs will eliminate weak-coverage areas.
New design approach
To nullify performance loss from interference, the researchers built on existing technology, choosing to focus on the following design principles.
- Personal network abstraction: Each user will have a virtual Wi-Fi network, plus the ability to define service set identifiers (SSIDs) and passwords, prioritize traffic, and add client devices. Users will be able to access their virtual Wi-Fi network as long as they are within RF range of one of the physical Wi-Fi routers or APs belonging to the domain.
- Sufficient infrastructure control: All physical Wi-Fi devices will be controlled (channel, power allocation, Wi-Fi properties, and which device a client associates with) from a central network controller.
- Overprovisioned physical infrastructure: Rather than be concerned about placement of Wi-Fi routers and APs, the researchers suggest having more devices than what are needed to insure coverage and redundancy. Devices not actively connected to clients are turned off to avoid interference.
The Stanford team call their novel approach BeHop. For the curious, BeHop is a combination of the jazz style bebop and the last "hop" of the network in the home.
The physical network consists of cheap consumer-grade Netgear APs reimaged with the OpenWRT and Open VSwitch. The research team also developed a Software-Defined Networking (SDN) controller. The paper's authors state: "Using our SDN controller, we can programmatically conﬁgure channel and power, add/remove clients, and handle WiFi management packets for discovery, authentication, and association."
The research team has started trials in Studio 5 of Stanford's Escondido Village. In The Stanford Daily article, reporter Alex Zivkovic writes that lead researcher Yiannis Yiakoumis said: "Those access points for the BeHop project support higher rates of connection than the existing equipment in Studio 5. In addition, the coverage is better for most students since the access point is directly in the students' room."
In that same article, Stanford graduate student and BeHop user Alexandros Manolakos stated he is finally getting bandwidth adequate for streaming content, and he was seeing up to a four times improvement in basic networking metrics.
BeHop is an interesting concept that builds on previous Wi-Fi networks that are centrally controlled. One of the better-known systems is Meraki Networks, a division of Cisco. During an email conversation with Yiakoumis, I asked how BeHop compared to Meraki. He replied:
"BeHop follows the Meraki approach in terms of a centralized controller. It goes one-step further as it exposes a single AP for every client, instead of all the physical APs that exist in the infrastructure. This way the infrastructure gets full control, and clients (phones laptops) don't have to bother about what is the best AP to connect to. We can force clients to use the 5GHz band, steer them to a neighboring apartment AP if it's closer to them, ensure proper channel allocation, etc."
What do you think of the Stanford team's solution to the Wi-Fi interference problem? Share your thoughts in the discussion.