Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Extreme's Purview draws application analytics from networks

Purview, Extreme Networks' network-based application analytics engine, is the winner of this month's SearchNetworking's Network Innovation Award.

Like it or not, application performance issues have become the domain of network managers. But fortunately, they no longer have to attempt to troubleshoot such performance problems with tools designed for software developers. An increasing number of network-based application performance tools have hit the market over the past year, and Extreme Networks took the industry one step further with Purview -- an application analytics platform that draws intelligence directly from any vendor's wired and wireless network devices without the need of additional infrastructure or agents.

Purview, the first product Extreme released after acquiring Enterasys last year, is the winner of this month's SearchNetworking's Network Innovation Award. Mike Leibovitz, director of mobility and applications at Extreme, explains the ins and outs of Purview.

How would you sum up Purview?

Mike Leibovitz: We refer to Purview as a network-powered application analytics engine. From a high level, what we're doing with Purview is harvesting traffic from the network itself into an analytics engine and ultimately visualizing on a dashboard what's transpiring on the network.

If you think about the analogy of a highway, for many years we -- collectively, as an industry -- built networks to go faster and faster. We'd see how many cars we could put on the road and we'd prioritize those cars, but nobody's ever really stopped to think about the context of what's happening in each one of those cars that's traveling down the highway. Purview basically opens up the door to those cars and takes a good look at who's in the car, what devices are in the car, what type of applications they're using and of course the location. We refer to that as 'context' or 'complete context' -- who you are, where you are and what applications and device you're using. And [in terms of] pattern matching, it's also being able to look at either specifics like an individual or … being able to understand what's happening at a macro level.

It gives you that capability to harvest information directly from the network and create business intelligence. It's unique because none of our competition has stopped really to think about leveraging information that's going all across the network.

There are a few third-party application performance management (APM) tools that use a network-based approach. What makes Purview different?

Leibovitz: One of the critical [differences] is we're doing this right from within the network; the switching infrastructure is providing information directly from the source. So by no means are we overlaying an infrastructure, and we're not taking any sort of performance hit.

On top of that, Purview is adding its information into our NetSight management tool, which is where all of this information is visualized. But where things get really interesting is NetSight has all of the information from the entire infrastructure, so it has all of the context -- users, devices, locations in the network -- and now the applications being used. We're definitely able to measure network performance and application performance from within the tool, similar to an APM or NPM [network performance management], but what those solutions don't provide typically is the full context that we're able to provide [because we're pulling that information] directly from the network infrastructure.

We're bringing the network to life without having to overlay any sort of infrastructure like in an NPM environment or put it at the perimeter of a network. With APMs typically you install software on a server. There are no clients required or anything like that.

Talk a little more about how this in formation is collected.

Leibovitz: In terms of user information, it depends on how the customer has configured their network -- user information and login attributes can be populated into the NetSight database to provide complete context. It's very much dependent on how the customer has deployed the solution. In terms of the authentication, that is coming back either from Active Directory or RADIUS, and it's being populated into our database as a single management platform.

It's very, very scalable because we're not taking a copy of the entire network's worth of traffic.
Mike Leibovitzdirector of mobility and applications, Extreme Networks

There's nothing that's really being collected in terms of personal information and, further, we're not actually collecting any data. We're not collecting payload information from traffic traversing the network; we're really just collecting the packet header, which would allow us to understand that somebody's watching Netflix, but we wouldn't necessarily know what they're watching. We wouldn't be able to say, 'Hey, you're watching Orange Is the New Black.' We could see somebody is using Instagram, but we wouldn't have the ability to recreate, show or expose the picture that you're uploading.

In the case of an Extreme network, the switching fabric is part of the production network; we can harvest directly from it. Any Extreme customer can start achieving application analytics by simply installing Purview into their network. In the case of a non-Extreme network, we would see installation of a flow appliance in a strategic part of the network.

We're doing it in a very scalable manner based on our architectural approach. We're able to do things at a very high rate of traffic. It's very, very scalable because we're not taking a copy of the entire network's worth of traffic. We don't need the data. We're not interested in the actual payload. We're just interested in the headers, and we're able to slice that off in our switches and send that over to Purview.

What can you achieve with the insight you get from network-based application analytics?

Leibovitz: It's very much [about using] context to create business intelligence -- using the network to get these patterns and being able to understand who's doing what on the network, potentially [to] optimize the network for better experiences and then potentially augment different aspects.

One [use case] that's become very popular in education, for example, is trying to get a good understanding of student engagement and being able to correlate that to student grades. A very powerful piece of information for a school, whether it's higher education or K-12, is being able to map and correlate the types of applications, the types of devices and the times [spent using them]. You put all that information together and can say, 'Look, the students who are doing the best in this particular school are spending two hours a day with an iPad in the library using a blackboard application.' Now if you can turn to a lower-performing school … you have better insights to make the better decisions.

But what about brick-and-mortar enterprises? What is the value of network-based application analytics when users are employees, not visitors or attendees?

Leibovitz: In the evolution of the product, we spent two years building it before we actually released it. We had some great beta customers -- education customers, healthcare, the New England Patriots at their stadium -- but we were also using it ourselves, with us being a pretty good example of an enterprise. Sure, we build hardware, but when you come to our offices, we represent just a regular enterprise.

We have what we call our 'inside sales reps,' who essentially do outbound calling to try to get leads for us. The manager of that team took a profile [of his team in Purview] and found out something very interesting. What he learned was his two best guys were spending three to four hours a day on LinkedIn, whereas most of his team was not. When he spoke to those people, he got a good understanding of what they were doing. They were doing this on their own accord; it wasn't part of a mandated program. And what they were doing was they were spending time on LinkedIn to learn about the people they were about to call -- what are they interested in, what are they posting articles about, where else have they worked? By calling somebody and being able to have a more human conversation, these guys were booking far more appointments than anyone else. Now on our inside sales team, basically everybody's mandated to spend time on LinkedIn. It's that insight our company learned, so as a result, we increased the number of leads and the number of appointments our inside sales team is booking.

All of these use cases sound very useful for CIOs, but what about the networking team?

Leibovitz: You spend so much time building a network and until it breaks, nobody really appreciates the fact you did a really good job and spent a long time maintaining this network. And when something breaks, it's 'the network's fault.' The first thing most people get attracted to with our product is being to answer this very question: Is it the network or the application? And because we have this built-in response-time recording, we can dissect that problem very quickly.

This was last published in July 2014

Dig Deeper on Network application performance

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.