Faultline doesn’t wait around. An article in our WiFi Wars series before the new year mentioned a Norwegian start-up called Domos (credit ASSIA for the introduction), an expert in executing WiFi optimization using machine learning techniques, and this week Domos discredited some of the big claims being made in the cutthroat WiFi software space.
An avalanche of WiFi announcements cascaded into inboxes this week, mostly from CES, so it was refreshing to sit back and get an account of what’s going on behind the scenes in the research world, from Domos’ CTO Magnus Olden. Domos may only be a 10-person team today, but the company has already claimed the scalps of Vodafone, Elisa, Telenor and Orange. Its contract with Orange Labs is an interesting one, which we touch on further down this piece.
While AirTies, ASSIA, SoftAtHome, Plume, and more have each cornered sizable slices of the market, Domos has repositioned away from what Olden describes as “classic WiFi management” by focusing on resource optimization on CPE.
“Typically, WiFi management aggregates data, then analyzes it, then outputs something like how likely a customer is to contact the call center, which I can tell you doesn’t work very well. What we do is optimize for a specific application outcome, whether it’s Netflix, Fortnite, Skype or any other application sending packets over WiFi,” continued Olden.
He broke down Domos’ technology into three core algorithms. Firstly, Domos App Classification detects apps in real-time based on a library of classified services. Next is the QED Probe, which measures the available networking resources from where it makes network quality predictions based on neural network models. QED means QoE delivered, based on the Broadband Forum’s standardized network quality metric framework (QED, MR-452) – which helps bring a level of clarity to the minefield of definitions for QoE in the WiFi industry.
Lastly, comes the Network Resource Orchestration algorithms, where resources are allocated to each application to optimize QoE for the home.
The Domos difference, according to Olden, is that while other WiFi optimization engineering outfits will claim to prioritize the most pivotal traffic over networks, they do not achieve this on an application-specific basis. “One of the great innovations we have succeeded in is randomly probing and then sending this data to the cloud for retraining neural networks. If an end user is playing Fortnite and the estimate is predicting this flow, then we prioritize that application’s traffic over everything else,” Olden explained.
“This is not about capacity, this is about orchestrating to achieve business benefits such as gaming first,” he added.
“Previously, we worked on optimization techniques like band steering or channel changing, but we decided to focus on what we are best at, which is machine learning and microservices with a standards-based approach,” continued Olden. “You need both band steering and optimization.”
Let’s take a look at a quick case study now. Domos recently received funds from Orange Labs and Telenor Sweden to train machine learning algorithms, with the goal of sniffing out hidden node issues in WiFi networks. The development will expand on the Broadband Forum’s work on the QED framework, with Orange and Telenor deploying probes on existing CPE, including SoftAtHome hardware, to achieve machine learning models capable of estimating network quality. Domos data suggests that hidden node issues are a major contributor to WiFi performance anomalies.
Olden could only tell us the project would take “some time” before results could be shared, or anything resembling a commercial deployment at either European operator could bear fruit.
While Domos has worked with SoftAtHome technology, Olden wanted to be clear that this wasn’t a commandeering attempt. “The scope of SoftAtHome does a lot more than us, from making the middleware and handling the full WiFi management, while we just want to do our one little thing,” he said.
This was a similar sentiment to that of Lifemote, a similar sized software supplier on the scene we spoke with last month. Domos believes strongly in a vendor-agnostic approach, where the importance of standards comes into play. “Standards are superheroes” – read the title of one of Olden’s slides – in homage to the brilliant work being done by the Broadband Forum, the prpl foundation and other non-profit industry bodies. This will be central to the progression of this market and wider associated markets, with Domos firmly in the camp that operators using unstandardized technology is a hindrance.
The way Domos sees it, the edge that ISPs have in WiFi management is their CPE and Olden noted how this is a strangely an under-used asset, which the vendor is taking advantage of, rather than taking a pure cloud-based approach. Domos assures us that no active analysis is done in the cloud so sensitive data never leaves the home – essentially saying that cloud infrastructure isn’t as essential for the WiFi management sector as some suggest.
We’ll throw in a smattering of data now for good measure. Domos notes that traffic from the device to the router is sent on rates significantly lower than in the opposite direction, from the router to the device. Test results found that from devices to the router, more than 70% of traffic was delivered at 21 Mbps or lower. Meanwhile, from the router to the devices, more than 70% of traffic was delivered at 39 Mbps or higher.
The point of all this, Domos concludes, is that different devices handle different environments differently – with some susceptible to noise, some to signal, and some to congestion. Olden highlights that fully explaining all this is difficult, and this is basically the thrust of what the vendor’s research is trying to figure out, with application resource optimization the end goal.
Another differentiator that springs out for Domos is its work on network slicing, a technique more associated with 5G networks than among the Domos customer target base, although we are aware of network slicing being used in WiFi networks. Olden explained that there are various flavors of network slicing and despite some of these offering benefits in fixing WiFi channels, Domos has stepped off the network slicing hype train for now – describing network slicing as low hanging fruit in the WiFi space.
On a final note, Olden expressed an alternative view on WiFi 6 to what we are used to hearing. “We don’t think WiFi 6 will be as revolutionary as people are marketing it. For us, the arrival of WiFi 6 makes things far more fun – but this is an evolution, not a revolution,” was Olden’s refreshing sign-off.