SDN in the Enterprise: It’s the Applications, Stupid
If virtualization has changed anything in the enterprise over the past decade, it’s that the capabilities of underlying infrastructure are no longer the prime determinant of the efficacy of data operations. Rather, productivity is driven by the extent to which discrete components, virtual or physical, can be harnessed toward a single, defined purpose.
In this way, it isn’t the SDN architecture itself that will prove the technology’s worth, but the networking applications the enterprise chooses to support and the way in which the enterprise implements and manages them.
Cisco seems to have taken this notion to heart and is already putting in place the framework for building and managing SDN applications. The company recently acquired a stake in Embrane, developer of the heleos network management stack, and is already working to bring key heleos elements into its SDN portfolio. These include IPsec site-to-site and remote access VPN services for beefed-up firewall capabilities, as well as advanced SSL-based load balancing and automated service provisioning. The goal, it appears, is to extend today’s top networking functions into the SDN realm. There, they can be applied to the dynamic network environments that will arise under the Application Centric Infrastructure (ACI) platform.
It is clear, though, that the challenge going forward will not be to simply mirror today’s static network capabilities in a software defined environment but rather to devise new solutions to the management problems unique to abstract networking. On-demand provisioning requires that network management tools be made available instantly as well, says Packet Design’s Cengiz Alaettinoglu, preferably with the ability to reach back to standard network infrastructure.You'll also need greater TLC for the controller, the repository for much of the network intelligence, plus extensive new visibility capabilities to keep an eye on all those new dynamic pathways and system relationships.
Ideally, the SDN management stack should do more than prioritize traffic and optimize packet flow, says networking consultant Mike Fratto. It should encapsulate various metrics like link utilization and switch queue depths and then map that data onto a physical topology. If SLAs are in jeopardy, the system should then be able to adjust loads on the fly and acquire additional resources from the cloud or wherever else they are available to ensure an appropriate user experience. This kind of functionality won’t be easy to program, but SDN makes it possible.
Of course, prioritization should not be discounted just because networks are dynamically configurable, says consultant Mike Finnerman. Under Unified Communications (UC), time-sensitive traffic like voice and video will still share bandwidth with standard data, even though they occupy different layers, and should receive special care when it comes to QoS, latency, jitter and other issues. Again, though, the good news with SDN is that these parameters can be established through automation. Applications can specify their unique needs to the controller, which can then compile the appropriate resources with little or no human involvement.
In classic fashion, SDN represents both a challenge and an opportunity. Layer 7 functionality is probably a little behind the curve right now, but, in all fairness, we’ve barely begun to explore how these environments operate in the real world, let alone how to control them.
But enterprise executives should know that even though the top vendors are positioning themselves along various architectural lines right now, the real differences will come in the networking applications they support.
Photo courtesy of Shutterstock.