Few enterprises can make the leap from traditional data architectures to nimble, virtualized operations all at once. So while leading vendors, and those of us in the trade press, continue to tout the wonders of software-defined this and that, many organizations are still struggling to maintain effective performance on bare-metal hardware.
Networking is particularly troublesome in this regard because most equipment has a longer lifecycle than either servers or storage. Plus, the connected nature of infrastructure makes it difficult to simply swap out components when they have served their purpose.
In fact, recent surveys suggest that it is networking more than anything else that is hampering the conversion to advanced architectures. Riverbed’s latest Future of Networking Global Survey, which canvassed more than 1,000 data executives around the world, found that adoption of even such run-of-the-mill developments like the cloud are being delayed because of a lack of advanced networking capabilities. While more than 90 percent of top decision-makers recognize the need for next-generation service delivery, a good 85 percent admit that legacy infrastructure is holding back their deployment plans.
One of the biggest hurdles in network conversion, says Alcatel-Lucent’s Joe Raccuglia, is the continued reliance on capex models for network upgrades rather than the more nimble opex approach of software-based service platforms. But just as SaaS has remade the productivity suite, so too can it bring the network more in line with today’s flexible, mobile data environment. With networking as a service (NaaS), the enterprise is able to lower its upfront costs and more accurately gauge the consumption of network resources to revenue-generating operations. At the same time, it provides ready access to the latest technological developments because NaaS providers are constantly trying to outdo one another in order to gain market share.
Implementation of advanced network architectures is certainly much easier in software than in hardware, but what is the best way to integrate a partially deployed SDN into a legacy network? Researchers from Technical University of Munich argue that the key challenge is designing a mixed control plane that functions smoothly with existing network management systems. This is a difficult undertaking because it must incorporate the vastly different ways that the two sides handle things like deployment, reconfiguration and measurement/monitoring. Without further research in this area, however, enterprises can expect only marginal results from their SDN deployments until nearly the entire network ecosystem has been upgraded.
This leaves the enterprise in a bit of a jam, however, because the competitive pressure to shed older networking equipment often runs headlong into the desire to leverage that gear until it is fully depreciated, and then some. But as The Register’s Dave Cartwright notes, there are plenty of ways to re-purpose older kit as more of the workload transitions to hybrid and abstract infrastructure. For one thing, even extremely old, sub-Gbps gear can still be used for a dedicated management LAN that connects the “lights out” adapter on the server to the management port on new network devices. If it is Gb-capable it can provide a dedicated Ethernet link for SAN traffic or function as a backup streaming LAN. It can even trunk multiple interconnects of an LACP or EtherChannel without affecting operations on a production switch.
Going from hardware- to software-defined networking is kind of like making the jump to light-speed in sci-fi: it’s best to do it quickly, but carefully. True, enterprises that are slow to implement agile networking run the risk of losing out in the service-driven economy to come, but those that jeopardize current workflows face an even greater risk in the economy we have right now.
Arthur Cole is a freelance journalist with more than 25 years’ experience covering enterprise IT, telecommunications and other high-tech industries.