Network management is evolving quickly. Today’s IT engineers are straining beneath mountains of data and an array of new data sources. How can they stay on top of it all?
According to Anand Srinivas, Nyansa’s CTO and co-founder, the answer lies in data analytics. In this interview, Srinivas opens up about how data analytics has changed the network operations function, how IoT devices will impact networks, and how analytics can help boost network performance.
SDxCentral: How is data analytics changing the way networks are managed?
Anand Srinivas: Big data network analytics lets you view and analyze more data across many more dimensions to gain insights that were not previously possible.
Vendor-specific management and monitoring tools along with manual CLI [command line interface] methods have been the traditional tools of choice. With network monitoring, you monitor the up-down status of your infrastructure (e.g., routers, switches), study generated graphs, and you generate alerts based on manually set thresholds, and so on. It’s a simple method that produces simple sets of data.
But this conventional approach doesn’t fit for modern networks. There have been major shifts in the market and technology. The biggest shift of all has been mobility.
We’ve gone from company-assigned devices, plugged into wired Ethernet jacks to smart phones and laptops connecting over a Wi-Fi network, accessing applications that may be hosted on the cloud, on-premise, or wherever. It’s a much more heterogeneous environment.
Just think about the amount of data traveling on modern networks and the host of things that could go wrong. You’re now dealing with various types of wireless devices, with different operating systems, different hardware versions, and a range of applications. IT engineers need to look at an extraordinary amount of data to gain any real, valuable insight into the client experience.
In traditional networks, engineers could fix user experience by looking at up-down status and studying a few graphs. Fixing problems was a matter of increasing speeds and feeds. Today, you need modern and intelligence web-scale analytics. There’s no way around it.
AI and machine learning provides new intelligence and automation that make analytics more valuable.
As larger amounts of data flood into the enterprise network, there’s no way to easily capture, analyze and correlate this data across the entire network to identify and remediate problems impacting device performance and user productivity. You need the solution to make proactive decisions and answer very complex questions. AI and machine learning make this possible.
One way to think about network analytics is in the analogy of consumer analytics. Think of Google, Amazon, or Netflix. The purpose of Netflix’s recommendation engine, for example, is to make suggestions to users that they didn’t even think about. Rather than relying on the user to know the questions they want to ask, the system makes recommendations as to what it thinks are the important questions. It sounds simple but it’s not and requires many different steps in a rigorous process.
The first step of network analytics is correlating, crunching, and indexing data. Historically, this involved manually reducing a complex dataset and comparing values to get answers.
But the next step is the big differentiator. Analytics systems need to make useful recommendations with a defined benefit. In the case of networking, they must quickly examine all incoming data from end to end and produce suggestions –– what we call actionable insights –– to network operations, network engineering or Help Desk staff. We can’t expect IT professionals to be data scientists: we need to translate the data into terms and actions they can use.
By combining AI, machine learning with cloud-based computing, petabytes of data impacting the client experience can be processed in the blink of an eye. AI and machine learning are then used to make recommendations, automatically generate models, and generate insights from the data that can’t be easily done by humans.
Now IT staff isn’t flying blind – not knowing what questions they should even be asking. The system does it for them. There are so many dimensions, so much data, so many things that can go wrong. It’s virtually impossible to look at everything. Pulling and analyzing server SYSLOG information, wireless LAN controller data, client traffic, WAN flows and application behavior is cost and time prohibitive.
And by sourcing all this data in the cloud, we can offer companies a comparative view into what other enterprises are doing across different aspects of their networks. It’s the first time any analytics provider has been able to give companies an industry-wide understanding of what is working and what isn’t.
What are some of the biggest challenges enterprise IT executives face today? And how is that different from a couple years ago?
IT executives need to transform enterprises from reactive to proactive workflows.
Modern access networks are inherently complex and when combined with the explosive growth of devices, data and apps running over the network, anything could go wrong, anywhere, at any time.
Many IT executives have thousands of sites around the world with tens of thousands of mobile clients and IoT devices that are accessing all kinds of different business-critical applications. Managing the performance of all these clients is a daunting challenge that conventional infrastructure management solutions just wasn’t built to tackle.
In the past, finding answers to site problems have been very anecdotal. Users complain about the network, and IT staff takes certain actions. Staff hope that complaints subside. But when you have lots of sites, lots of data and lots of different devices, this anecdotal approach doesn’t work well. Problems aren’t reported immediately or at all, it’s difficult to quantify the impact of changes to the network, and the amount of data in play is immense. Intelligent analytics is the ideal solution.
An analytics system crunches all that data and quantifies both the experience and the health of client devices on the network across all these different sites. It then figures out where the problems are hiding and tells IT staff the actual benefit derived if fixed, such as the number of lost hours of client connectivity that can be recovered, before users even know there’s some sort of problem. This allows IT staff to become proactive and know exactly where, when and why to best invest time and money.
Data analytics are supposed to make enterprise IT teams more proactive. Please give an example of how data analytics helped find and fix a problem for one of your customers.
One of our large retail customers saw on our platform that users were experiencing slow Web performance. The users just thought the Wi-Fi network wasn’t working properly. That simply wasn’t the case.
With data analytics, they were able to automatically baseline how and when these performance issues were occurring and what changes were made to the network at the time that might have caused the problems. The customer was able to see that during that time, a network design change was made that forced traffic to traverse two different firewalls. This was a vendor issue that hadn’t been communicated to the customer. When they fixed the problem, they saw a three-fold improvement in client Web performance. To find and fix this problem without such a system would have taken days or even weeks to determine.
One key factor in this process that can’t be understated is feedback. If a change is made to the network, what was the impact of that change? Because network analytics constantly stares at and trends the behavior of the network from a client experience perspective, enterprises can quantifiably prove that a given change had the intended effect with a clear return on investment.
IoT is creating a lot of challenges for IT departments. Can you address some of those challenges?
What we’ve seen in our customer environments is that IoT devices are often purpose-built for specific business operations functions, such as infusion pumps at a hospital or IoT robots on the production floor, unlike more general-purpose devices (e.g., laptops, tablets, smartphones). IoT devices can also potentially help with the line of business, in certain verticals. With that distinction in mind, we see a couple of important aspects of IoT that create challenges for IT departments.
1) Does the IoT device actually work on the network, and does it actually accomplish its purpose? Is the robot on an assembly line able to connect to the network and perform its activity? When you think of client experience for an IoT device, the first step is to make sure it can get on the network and do its job. One of the challenges is that manufacturers of IoT devices sometimes don’t pay much attention to the networking side, incorporating a subpar networking stack that results in erratic behavior with different parts of the network. This is a massive problem.
IT departments often don’t know what IoT devices are even operating on the network! So how do you get a sense of all these devices that, because of their quirkiness, might end up creating problems?
2) Beyond basic visibility, what are the various IoT devices actually doing on the network? Who are these devices talking to? How much traffic are they sending? Is that normal or abnormal? How can we best secure them?
Network analytics gives IT departments insight into the performance and health of all IoT devices operating on the network so they can ensure that devices are fulfilling their function and that they are behaving “normally”. The performance of IoT devices can directly affect an enterprise’s bottom line: when there’s a network-related stoppage on a manufacturing line, that can translate into millions of dollars in lost revenue. When IoT devices are mission-critical, it’s more important now than ever for enterprises to invest in network analytics to ensure future growth.
How does data analytics help improve the performance of devices on networks?
That’s precisely what it is designed to do. But keep in mind that client performance needs to be understood and analyzed across the entire network stack – from the client devices accessing the network all the way to the destination application and everything in between.
There are millions of client transactions that must be analyzed. How are mobile devices connecting to wireless? How are they connecting to the network? How are they authenticating? How are they being assigned an IP address? How are they crossing the wide area network (WAN)? And finally, how are they behaving with applications? To really gain useful insight and answers to these questions, you need a broad range of different data sources that feed into an end-to-end analytics platform.
This includes traffic flows from routers, IoT telemetry, wireless data from WLAN controllers, RF metrics from clients, wired packets, SYSLOG messages from servers and so on. Sound overwhelming? It is.
Ingesting all these sources, analyzing and correlating them, tracking baseline to determine when things are good and bad, generating alerts, and then making recommendations that can be acted upon is the key to how analytics drives better performance in modern enterprise networks.