Wi-Fi analytics get real

A number of Wi-Fi analytics tools have been brought to market over the past few years, and while most organizations have yet to dip their toes in the Wi-Fi analytics waters, our research shows that those who have are realizing significant benefits.

To be sure, there’s still lots of room for innovation in Wi-Fi analytics, but the glowing reviews we’ve received in our interviews with network practitioners across a wide variety of industries and applications indicate explosive growth ahead over the next few years.

While the two terms are sometimes used interchangeably, it’s important to understand how analytics is different from analysis. Analytics applies “what-if” thinking to very large problems that would otherwise be intractable. Whereas analysis is appropriate for bounded, well-understood problems, analytics can be applied in situations where you don’t know what you’re looking for – and can do so with the speed and flexibility that is essential in network operations today and, in most cases, beyond the capabilities of human information processing alone.

It’s easy to see how analytics can be applied to Wi-Fi networks. While suspect behavior almost always triggers an exploration of the data available, the sheer volume of information, even in mid-sized networks, generally precludes a quick conclusion. The quickest path to extracting meaning and value from all that data, and setting upon the optimal path to a solution to what might seem like a simple problem at the time – but which often morphs into very long weekends looking through logs and settings – is the right analytics tool.

As we learned from our conversations with those already putting Wi-Fi analytics to work, the solution to any given problem often leads well beyond the common and the obvious, and with the constant pressure to minimize time-to-solution so as to keep networks on the air today always motivating innovative productivity-enhancing strategies and solutions, Wi-Fi analytics brings exactly what’s needed to the table.

Widespread knowledge and understanding of the technologies and strategies of network analytics today remains limited, as is to be expected in any new field of technical endeavor. So, in an effort to close this knowledge gap, we decided to take a look at how an investment in acquiring, learning, and applying Wi-Fi analytics tools is benefiting operations teams.

We interviewed IT professionals at four different organizations, asking three fundamental questions: Why did you make the investment in network analytics? What problems are you solving that couldn’t be addressed before? And what additional network analytics capabilities would you like to have going forward?

Here’s what we learned.

University of Washington troubleshoots Wi-Fi

David Morton is director, networks and telecommunications, for the University of Washington. David and the IT Infrastructure team are responsible for wired and wireless networks at the university, plus university hospitals, the state’s K-20 school networks, and more. The network includes 12,000 APs and serves more than 80,000 unique users per day and a few hundred thousand devices connecting per week.

Morton noted that as his networks continued to grow in scale and complexity, more issues and problems were appearing than were being reported. With Wi-Fi access becoming mission-critical, he needed better methods for finding and fixing problems, proactively if possible.

He chose Aruba’s NetInsight analytics tools to help him identify specific problems and recommend solutions.

For example, a “clicker” application that enabled students to answer questions in class was not working properly, but with only 40-50 out of 400 students in larger classrooms being affected, identifying the cause was very difficult. Multiple classrooms had the same problem, but the affected user base varied without a pattern.

A check of the management console revealed no obvious issues, so Morton used NetInsight to analyze across clients, controllers and destinations. He discovered that the issue was multi-faceted: a Wi-Fi configuration change was indeed required to balance users across radio channels, but the primary cause was localized to a cloud-based third-party software service provider, which required additional scale in their infrastructure along with bug fixes.

“We couldn’t have found this without [NetInsight],” Morton said, “and now that [NetInsight] knows about the problem, it will automatically report if it recurs again – at the UW, or even another [NetInsight] site.” Morton finds great value in the tools’ cloud-based, multi-client nature.

Going forward, Morton wants to be able to unleash the tool to find actionable insights from data already present within the network, and to automate responses via management-console feedback. He’s also looking for more nuanced intelligence as networks continue to increase in capacity.

NVIDIA seeks a holistic network view

John Kilpatrick is senior network architect at NVIDIA, a company that specializes in graphics, AI and high-performance computing. His primary responsibilities include wireless and the enterprise access layer of the firm’s network. NVIDIA has 15,000-20,000 global concurrent clients and 3,000 APs, and they use Nyansa Voyance for analytics.

“We’re making a very significant investment in improving wireless infrastructure,” Kilpatrick said. “We regularly ask ourselves what can we do to ensure the best quality of experience for our users. So, we need an analytical, data-driven approach to quantifying the user experience in terms of throughput, capacity and coverage.”

Kilpatrick said he has seen a broad range of problems in network operations over the years, including in such functions as DHCP, authentication and onboarding. “It was difficult to see the nature of a given problem before we installed Nyansa, and the problems weren’t always wireless. These could include a bad browser version, unpatched security flaws, a slow DHCP server and many more.”

Installing the Nyansa software enabled a more holistic view of the network. “I use it every day, and it makes my job so much easier.”

His ongoing goal is to know about problems before users are affected. And he’s interested in enhancements that enable more operational data to be integrated into his analytics solution, such as Cisco Service Assurance, software agents on clients, and integration with enterprise mobility management (EMM) solutions. Standards and APIs for analytics are also of interest at NVIDIA.

Optix Media improves performance

Shane Moulton is president of Optix Media, LLC, a managed ISP with a focus on student apartment complexes. The firm provides a turnkey Wi-Fi service, emphasizing customer service, partnerships and long-term relationships with property developers and managers. The scale of operations is national, with 6,000 APs and around 120,000 connected devices.

Students can be a very demanding audience. “Our customers expect perfection,” Moulton said. In an effort to meet that goal, Optix Media is using the analytics capabilities built into the Mist Systems management software.

“We originally got involved with Mist because we view them as an innovator, but not specifically for their analytics,” Shane said. Optix Media deployed the Mist solution at their biggest pain point, a residential complex with 900 students.

The use of analytics has indeed proved quite valuable. “We made a change to the network and found a DHCP error right away,” Shane said. “We were impressed with the very rapid time to solution, much better than we’ve ever seen in the past.”

Analytics is now essential for all new deployments, having been used to resolve other problems such as client driver issues, wireless incompatibilities, and older devices causing slow performance.

Looking ahead, Shane said the “software needs to be as smart as the tech support people,” a direction he mentioned Mist is pursuing. “We need to have wireless tools that are as good as those on wire, and analytics must be as integrated as possible into every Wi-Fi network.”

University of Mount Union moves toward a self-managed network

Tina Stuchell is executive director of Information Technology, and Dave Smith is assistant director of IT for Technical Services at the University of Mount Union, in Alliance, Ohio, a school with about 2,200 full-time-equivalent students, each using three to five devices, often simultaneously. The school operates more than 1,300 APs, including a stadium with guest access.

The school’s objective is to cover the entire campus with high Wi-Fi density to ensure sufficient capacity for their demanding student audience. They bought the Extreme Analytics tool when it was introduced a few years ago as one of the first analytics products to hit the market.

“Our goal always is to optimize user experience,” Smith said. “We’re always asking ourselves how we can further improve our operations. We need more productive troubleshooting and more efficient operations, and the improved visibility that comes from having a great analytics capability helps us in being proactive.” He added that they installed analytics during a major network refresh, an opportune time to explore new innovations.

As an example of the value of analytics, Stuchell and Smith mentioned a problem they were experiencing in application latency – specifically, why Netflix would occasionally be buffering or dropping connections.

“Is the problem inside the campus network, at our ISP or at Netflix itself? Given the massive amount of data to plow through, analytics has proven itself essential in dealing with problems like these. We also like having a customizable dashboard, the ability to identify malware and spam-related issues quickly, and being able to do rapid traffic-pattern analysis alone more than justifies the investment,” Stuchell said.  “We’ve improved time-to-solution and our ability to be proactive well in advance of any impact on our user community.”

Looking ahead, “the graphic displays and monitoring are helpful and informative, but analytics tools in the future need to do more than just provide logs, graphs and charts,” Stuchell said. “They need to alert us before the end-user experiences a problem and give us exact steps to fix the problem – or, better yet, automatically address the problem even if working with other systems such as policy, DHCP, etc., is required.”

Smith added that he isn’t quite ready for analytics to, for example, shut off a switch port, but such an action with operator-in-the-loop approval would be fine.

“Networks just need to run, with little to no operator intervention,” Stuchell said. “They really need to be as self-managing as possible, making changes automatically. Analytics moves us towards that goal.”

What vendors need to do

Analytics is adding significant value in the shops we spoke with, and undoubtedly many more. As we learned from our conversations with users, though, there are a number of issues and opportunities that require action on the part of the vendor community:

• Standards – There’s a need for more open access to raw data from vendors. Knowing where the data is and how to interpret that data is critical. Open standards for device-specific access to operational data would certainly help, but developing them will be daunting without appropriate standards and vendor cooperation.
• Data completeness and validity – A major concern is validating the integrity, completeness, and interpretation of the data points involved. That old saying, “Garbage in, garbage out,” still applies, and automated analytics-driven decision-making will be impossible without this matter being resolved.
• Automation – We also see a clear need for a closed-loop feedback mechanism between analytics and settings within the management console. As analytics evolves from reactive to proactive, the focus will shift from what’s happening to what might happen and perhaps even will happen, a scenario we find realistic given advances in machine learning. Visual reporting will remain vital, but automatic actions are clearly within the realm of possibility – even probability, as networks become too complex for real-time human-in-the-loop management. Standards will also play a big role here.

The gradual transition to software-defined networking (SDN) now underway presents a very broad opportunity to address many if not all of the above.

There’s also a clear emphasis on the importance of cloud-based implementations and even multi-client cloud-services implementations of analytics. This is an exciting opportunity in that patterns of behavior noted at one client might indicate at least the possibility of a similar or related challenge at another – and often well before operations teams or end-users might otherwise take notice. There is, of course, the potential for new privacy and security concerns arising in this context, but it’s very likely that acceptable solutions to these challenges can be found.

Due to the size, scale, and the mission-critical nature of the majority of Wi-Fi networks installed today, it’s clear that ever-more powerful analytics tools are well on their way to a position of prominence within the arsenals of network operators everywhere.

We are especially intrigued with the possibility of enhanced machine-learning-based automation via a feedback loop between analytics and management consoles. This could even extend to, for example, operations-policy, security and overall network upgrades being driven by analytics.

As Wi-Fi analytics seems destined to evolve from a separate function and ultimately disappear into the network management woodwork, the benefits we’ve just begun to explore will clearly become even more visible – and valuable – in the years to come.