SDN is widely viewed as virtualization of the last refuge of hardware-based enterprise infrastructure: the network. This is not wrong, but some would argue that it is not entirely accurate.
To be clear, virtualization on the network level is not the same as virtualization in the server farm. While they both involve adding a layer of abstraction on top of static hardware resources, server infrastructure is likely to see much improved resource utilization and operational efficiency by virtue of the fact that processor utilization and efficiency were so poor to begin with.
When it comes to the network, however, the real benefit is flexibility in the allocation of resources. Now that entire network architectures can be configured with little or no changes to hardware at all, the notion of customizing the network for user, application and even data needs is no longer the stuff of science fiction. Indeed, this is why the market for SDN seems so bright. According to Grand View Research, global SDN sales are expected to near $5 billion by 2020.
Still, many people make the mistake of viewing SDN as a purely networking phenomenon. As Mellanox’s Kevin Deierling points out, the change is much more fundamental. Going forward, the network can no longer been seen as a means to connect discrete components of the enterprise infrastructure. Rather, we now have the makings of an integrated data platform capable of supporting a fully elastic and eminently scalable data workload engine. In other words, the network is no longer a thing to be built and managed in support of server and storage operations, but an integral piece of a holistic “compu-storage” entity designed to accommodate the dynamic needs of modern data environments.
What’s the difference between that and what we have today? In practical terms, it means that issues like bottlenecks and latency are no longer viewed from a network perspective. In the virtual future, the overriding issue will be to enhance performance, not simply locate and repair network bottlenecks or latency issues. As EY America’s David Nichols points out, the new digital business model will be heavy on communications and collaboration, which will rely more on connectivity and connectedness than shoring up the capabilities of any one computing resource. It’s the difference between simply changing the air filter to boost engine performance and performing a complete tune-up.
Some people may already recognize this as the “software defined data center,” but that is only part of the equation. The other is the advent of modular infrastructure that will be more apt to handle the vagaries of the SDDC, says IDC’s Glen Duncan. In a modular setting, you don’t have network or server and storage issues. You have an issue with a particular computing module. If performance is lacking, the problem is tracked to a module, which is then replaced. In many ways, this is a much simpler way to build, repair and manage infrastructure because much of the system testing, integration and configuration will either be handled by automation or fade away entirely.
In this vein, SDN is only a networking development when viewed through the lens of today’s server/storage/networking triarchy. But like the PC revolution of the 1980s and the Internet revolution of the 1990s, the dynamic data environment represents an entirely new way of interacting with the digital universe.
Virtualization of network infrastructure is the last piece of the puzzle. The now-complete picture is more amazing the most of us had realized.
Photo courtesy of Shutterstock.