The following is a Q&A from Mojo Networks with the company's CEO Rick Wilmer.

Q: Where are we right now in WiFi? What does a traditionally-structured WiFi setup look like, and where does the controller “live” in such an arrangement?

A: Enterprise WiFi is going through a few important transitions at the moment, all of which have the potential to disrupt the market and vendor landscape. For large campus environments, typical deployments have all been controller-based. A controller-based WiFi network has the APs connected to the controller for the management plane, control plane and data plane functions. These controllers are network devices placed within the customer’s network, usually located in medium to large MDFs or data centers, aggregating all Wi-Fi traffic to the distribution or core. However, as controllers are networking devices, various functions are typically offloaded to other product suites, usually for fairly substantial costs. These may include network management systems, security systems or guest management systems, which are added to “complete” the overall feature set requirements of today’s advanced enterprise WLANs.  These additional systems usually come in the form of ancillary servers that are typically located in the core data centers, sometimes in the same locations as the controllers.  The big controller-based vendors also require separate appliances to add even more features, such as location and analytics, as well as Wireless Intrusion Prevention (WIPS).

Q: What does the future WiFi setup look like? Why is getting rid of the controller important and what replaces its ability to manage WiFi access points?

A: The future WiFi setup eliminates the physical controller and replaces that functionality primarily with software running in the cloud. In some architectures, some of the work traditionally done by the controller is also moved to the access points. (Note: access point chipsets have advanced tremendously since controllers were first invented and now have enough processing horse power to do work not possible in the past.)

The shift to the cloud is inevitable because the advantages are so profound. It’s just a matter of when and who will drive the change—a new upstart vendor or a large incumbent.

There are two categories of major benefit. The first I deem as “operational” benefits, which have been the basis of the value proposition for the first wave of “controller-less” deployments. These include advantages like the savings that come from the simple absence of a controller, in terms of space, energy consumption, licensing, etc. Cloud-managed WiFi also provides businesses with flexibility and better agility. If the demand for a specific WiFi feature or service increases, it’s easy to scale up cloud resources immediately to meet the need. You cannot do this with traditional controller-managed WiFi. Cloud WiFi also enables rapid enterprise-wide software updates, which is particularly critical when it comes to Wi-Fi security. There are no controllers that need to be updated individually. Customers can roll out regular software updates with ease, including security updates. These upgrades can also happen on a per-AP basis, so that the entire controller is not taken down (and thus all of the APs) during a code upgrade.  All of the functionality you need to buy as separate appliances and SKUs with controllers (management, security, analytics, location, etc.) are typically included with a cloud solution. Cloud WiFi cuts down the high cost of hardware. In the old controller model, an enterprise WiFi project was a massive capital expenditure, significantly driven by the controllers and associated appliances.

The second benefit is even more significant. With a controller, you are limited to the processing and storage resources within the controller. With the cloud, you effectively have access to unlimited compute and storage. When put to use intelligently, the cloud can deliver capabilities that no controller can ever match. These capabilities are based on three primary core technologies: big data, machine learning and AI. Together, these technologies, running in the cloud, can make WiFi work better.

I will step back a bit to help to frame the importance of this. It’s well understood that WiFi is an increasingly popular way to connect to the digital world. It is pervasive in the workplace and many devices used for mission-critical work do not even have a wired port. In some cases, we’ve seen customers literally turn all of their Ethernet ports dark and move to WiFi as the exclusive connection method. Microsoft, for example, has disclosed this publicly. Under this scenario, two things happen: 1) huge savings are generated by decommissioning the wired network 2) the wireless network now becomes truly mission-critical.   

We believe you cannot make a WiFi network reliable enough to be the exclusive network unless it’s managed by the cloud. The reason? A Wi-Fi network, unlike a static wired network, is a dynamic, constantly changing environment. The storage and compute resources available in the cloud are necessary to control this environment on a real time basis. A truly intelligent, high-performing, self-healing network requires big data, machine learning, and AI in order to continuously monitor Key Performance Indicators (KPIs) and to convert them into essential, actionable insights for network administrators.

Q: Why is open-source important to the future WiFi model?

A: Customers are rejecting the way the networking industry has operated, which has been a proprietary closed system. The shift toward more flexible systems is already happening. IDC is predicting significant growth in the cloud wireless segment: by 2020, they expect it to represent 34.3 percent of the overall market.

While I can’t claim to be a tech historian, I am unable to recall a customer driven revolt in any technology sector like that which has occurred in networking over the last few years. Perhaps the advent of Linux may be the closest comparison. For a revolt to happen, there first need to be grievances that have been ignored for so long that people begin to feel that conditions are unbearable.  These grievances lead to actions, which may include the formation of protest groups. Finally, there is a spark, which triggers the actual rebellion. In networking, the problems that result from proprietary solutions have resulted in deep grievances. These problems include, among other things, high pricing, vendor lock-in, slow innovation, poor quality (e.g. buggy software), extra charges for many things that should have been included in the price of the product (e.g. the courtesy of receiving support), and duplicitous offerings couched under the term “professional services.”

Frustrated customers have taken action, which I find remarkable. This action has come in the form of a number of open networking groups, such as Open Compute Project, Open Network Users Group, and the Open Network Foundation. The catalyst for the revolt being the advent of SDN or NFV or whatever label you want to attach to a world where software (often open sourced) defines functionality and hardware becomes standardized and interoperable. The mission statements of these groups directly address their grievances, which, I believe, are largely a result of how the networking industry has comported itself for way too long.

Open Network Users Group (ONUG): The ONUG Mission is to enable greater choice and options for IT business leaders by advocating for open interoperable hardware and software-defined infrastructure solutions that span across the entire IT stack, all in an effort to create business value.

Open Network Foundation (ONF): We are an open, collaborative, community of communities.  The ONF serves as the umbrella for a number of projects building solutions by leveraging network disaggregation, white box economics, open source software and software defined standards to revolutionize the carrier industry.

Open Compute Project (OCP): The Open Compute Project Foundation provides a structure in which individuals and organizations can share their intellectual property with others and encourage the IT industry to evolve. In designing commodity hardware that is more efficient, flexible, and scalable, we’re redefining tech infrastructure. Together, we’re throwing off the shackles of proprietary, one-size-fits-all gear.

One analyst recently said “the general networking vendor is an old business model.” I tend to think about it a little differently. I believe “the value system and business model of the old networking vendor does not meet the needs of today’s customer.” The networking industry needs to change and we should advocate and do all we can to drive that change.

Q: How will the adoption of newer WiFi standards affect management technology?

A: Controllers are a bottleneck in the WiFi data path today and the bottleneck will be more severe with adoption of newer WiFi standards. WiFi link speed has increased by three orders of magnitude from the original 802.11 standard (2 Mbps) to 802.11ac Wave2 (1.7 Gbps) since the standard was launched. The number of supported APs per controller has grown from 50 to 5000. Over the same time period, uplink throughput of most controllers has only increased from dual 1 Gbps to dual 10 Gbps. WiFi throughput improvements have already outpaced the aggregate uplink capacity of Controllers. Today’s controllers are only capable of switching up to 20-40 Gbps of aggregate traffic upstream, while APs at peak network throughput can pump 1+ Terabyte of traffic on a fully loaded controller. The upcoming 802.11ax standard promises to increase network throughput by a significant amount. The gap between the traffic generation capacity of APs and the switching capacity of controllers will continue to widen in the future. Controllers will not be able to keep up.

The best approach to solve this is with a cloud architecture that decouples the data path from management plane. Peak throughput can be achieved (without incurring controller’s cost) by bridging WiFi traffic to the L2 network, or tunneling this traffic only when and where it makes sense. (i.e., guest traffic to a DMZ)

A centralized management plane in the cloud is highly scalable and operationally easier to manage, while a distributed control plane is infinitely scalable and resilient to failures. In 802.11ax, performance improvement comes from how traffic and radio characteristics are managed using the new hooks in the standard, unlike link speed improvements of the past standards. Big data, machine learning and AI can help achieve optimal operating configurations of these new hooks, and dynamically adjust them to achieve the best performance possible. And this can only be done with the cloud.