alphaspirit - Fotolia


SDN analytics offers key to smarter, more adaptive networks

SDN analytics promises to make software-defined networks smarter than ever, culling insights from vast amounts of big data and updating operations accordingly.

SDN is often described by networking professionals as the separation of the control and data planes, and while technically correct, this rather clinical statement doesn't even hint at the actual benefits of SDN. A more interesting and meaningful definition of software-defined networking is the incorporation of policy into the definition, architecture, implementation and operation of the network. So, if the network is indeed the circulatory system of the organization, SDN makes this plumbing smart, flexible and adaptable. SDN analytics can make it even more so.

Software-defined networks are dynamic, responsive

Smart is today, of course, essential. Networks have become truly mission-critical and so complex as to require a good degree of intelligence -- human and otherwise -- in operation. This is the real beauty of SDN -- provisioning new intelligence into what used to be a relatively simple collection of dumb pipes, with the key benefit of enhancing performance, security, traffic management and resilience. The specification of operating policies and their enforcement via the mechanisms of SDN dramatically enhances the nature of the network, via the ability to modify operational behaviors based on changing conditions. While management consoles remain important in the process, a degree of automation -- the extent of which depends upon implementation specifics -- is introduced. In other words, software-defined networks can change their behavior in accordance with policy and based on conditions that arise dynamically in the course of normal operations.

The key to solving any computational problem is to understand the meaning that the available data represents.

The big question is, of course, how to implement this behavior. How can a potentially vast array of operating conditions and events be quantified in such a way so as to enable SDN to respond accordingly and appropriately? Indeed, we will often see a "more variables than equations" challenge, wherein the "right" answer is often anything but obvious. Is the network under attack? What's the best way to respond to the failure of a particular network element? Suppose a software or firmware upgrade results in anomalous and even uncorrelated adverse behavior? A few taps on the management console aren't likely to fix the problem.

The key to solving any computational problem is to understand the meaning that the available data represents. Network logs and operational databases contain a potentially vast array of interesting information, but few humans (read: likely none) have the ability to search, correlate and, again, understand this information without appropriate tools. And, as it turns out, such tools have been fixtures in many domains of scientific computing for decades; they fall under the general heading of analytics.

SDN analytics offers new insights

Like SDN, the term analytics has a number of definitions, but one that's particularly useful here is: "The set of tools and techniques applied when you don't know what you're looking for." In other words, "analytics" indicates analysis strategies and techniques applied to big data; in SDN analytics, that includes all of the big data at the heart of networking today. While a human -- especially one under pressure during a crisis -- would usually have a hard time sifting through operational data while looking for patterns that illuminate the root cause of a particular issue, analytics tools are designed to do exactly that. SDN analytics apply both numerical and graphical (visualization) techniques to offer insight as quickly as possible.

So, what happens when we meld the power of analytics, especially on an ongoing, as opposed to ad hoc, basis, to the flexibility inherent in SDN? We get the network of the future -- self-monitoring, self-optimizing, self-correcting and always in sync with policy, thanks to a feedback loop between continual SDN analytics and operations. This form of automation is of course a branch of artificial intelligence, and, thus, there's a lot to be both attempted and proven before it becomes common. But imagine a cloud-based, cross-organization analytical capability that alerts one operations group about potential issues based on the experiences of another. Imagine the management console of the future, automating the response to conditions that might have previously resulted in performance degradation or outright failure. Imagine the intelligence of a software-defined network, optimized through the ongoing analysis of operational data. Consider the potential benefits of improvements to network planning and operations staff productivity, reductions in operating expense, and enhancements to overall reliability.

One might even argue that the software-defined network's promise cannot be realized until we recognize SDN analytics as one of its essential elements. In fact, consider that argument made.

Next Steps

Where big data and SDN meet

SDN makes DPI technology more effective

How does SDN enable automation?

This was last published in February 2016

Dig Deeper on Network automation and intent-based networking