In the SDN Era, the Edge Needs Love Too

Software defined networks and the distributed data center will require intelligence, interoperability, and agility at the network edge.

By Arthur Cole | Posted Nov 22, 2013
Page of   |  Back to Page 1
Print ArticleEmail Article
  • Share on Facebook
  • Share on Twitter
  • Share on LinkedIn

With the growth of Big Data, mobility, and machine-to-machine (M2M) communications, networks are increasingly under the gun to handle large volumes of data amid an ever-shifting array of consumer, data center, and cloud-based resources. In response, the enterprise has turned to advanced network fabric technologies and software defined architectures to build greater flexibility into their communications infrastructure.

Unfortunately, most of this activity has taken place on core infrastructure. But as more and more organizations now realize, the edge needs some TLC as well, preferably with higher degrees of intelligence and interoperability so it can take on a more sophisticated role than simply the place where you put the firewall. The edge, in fact, is quickly turning into the brains of the enterprise network. It decides where and how data gets shuttled through internal infrastructure to both reduce the demand on network resources and improve response times.

New edge routers like Juniper’s MX2020 can convert the edge into a rapid service delivery platform that in some cases can improve performance by 70 percent, even for rich media content like HD video. The MX2020 utilizes the JunosV App Engine to provide specialized gateways to and from service provisioning resources while still enabling massive scale-out capabilities that are a vital component of advanced cloud architectures.

Increasing the interoperability of multiple edge devices also goes a long way toward matching the dynamic networking capabilities now permeating core infrastructure. To that end, Extreme Networks has devised the Open Fabric Edge platform as an extension to the existing Open Fabric system that provides automation, intelligence and open standards networking to internal infrastructure. In this way, the enterprise can utilize the ExtremeXOS software to enable an integrated, end-to-end virtual network that can be used to provide full IT-as-a-Service functionality across dynamic configurations of pooled resources. At the same time, it allows the enterprise to do away with multiple discrete networks for, say, Unified Communications (UC) and video delivery.

This dance will only become more difficult as virtualization continues to upend the longstanding relationships between network elements, says QuoCirca’s Rob Bamforth. Expect to deploy increasing amounts of intelligence both on the edge and at the core as software defined paradigms produce ever-more convoluted network architectures. In fact, "the network" itself will become something of a misnomer as architectures evolve into multi-layered platforms of connectivity services, each with their own protocols, performance metrics and policy configurations. A crucial component of this increased intelligence will be real-time data assessment to predict future usage patterns or, at the very least, to calculate the effect of current demands on network resources.

And let’s not overlook the fact that the edge itself is becoming increasingly less defined as things progress, says Apigee’s Michel Burger. As data environments become more distributed across colocation services and the cloud, the challenge of matching user requirements to available resources will become more complex. This is why concepts like elastic provisioning using application-based APIs are gaining stature. With the app self-directing itself through a maze of centralized and distributed resources, guided by robust analytics that can gauge resource configuration and availability along the way, the concept of the fully automated, globally distributed data center takes a step closer to reality.

It is conceivable that at some point, concepts like core and edge will become archaic. The network will be whatever software architecture or hardware device a given packet of data encounters at any time. This could be a traditional switch or router, a software defined virtual device, or even a user handset. In the meantime, however, the enterprise would do well to take a hard look at the discrepancies arising within its own physical infrastructure. The best way to make the most of newly virtualized core systems is to push some of today's networking advances to the edge.

Photo courtesy of Shutterstock.

Comment and Contribute
(Maximum characters: 1200). You have
characters left.
Get the Latest Scoop with Enterprise Networking Planet Newsletter