Tracking the Elusive Definition of 'Network Fabric'
When an entirely new datacenter architecture like the cloud comes along, it's only natural for smart people to ask, "What is this thing?"
But while a single definition that satisfies everyone has proven elusive, we at least have to give the IT industry credit for trying. What's odd, though, is that many of the cloud's enabling technologies have not undergone this same process. Concepts like the network fabric seemed to suddenly enter the IT lexicon with little or no discussion as to what they mean and how they are to be implemented.
This isn't to say that the idea itself is difficult to understand. Fabrics, after all, are simply a flattened, federated network architecture that replaces the old point-to-point relationships of the past with much more dynamic and scalable multipoint-to-multipoint architectures. Along that way, you greatly enhance data throughput and flexibility while reducing capital-intensive physical infrastructure and redundant layers of management and manpower.
Sounds simple. And in truth, it is -- once you have a working fabric firmly in place. The journey from here to there, however, is anything but.
Aside from all the thorny issues of incorporating legacy infrastructure into the mix (unless you want to chuck your entire plant and start over), there are a range of issues governing traffic management, protocol support, I/O queuing and the like. If you thought Spanning Tree was complicated, you ain't seen nothing yet.
It's only natural, then, that top network vendors like Cisco, Brocade and Juniper are pouring a lot of resources into fabric development and have turned their vast marketing machines not only onto selling the fabric concept but in guiding the transition to a flatter network environment. At the same time, network fabric technology has taken center stage at many of the leading cloud computing conferences and seminars around the world. A nod to the fact that without a high degree of network flexibility none of the many and much vaunted cloud platforms would get off the ground.
Further complicating matters is the increasing number of fabric solutions hitting the channel. For every Cisco and Brocade out there, there are numerous smaller developers touting advanced switching and/or OS and management stacks all providing the highest densities, most flexibility, lowest latency and easiest installation, etc. -- in short, the true fabric technology that will best support enterprise endeavors into the cloud and beyond. And in a broad ripple effect, fabric technology is starting to influence development of support technologies like traffic management appliances and network clustering tools.
So where does all this leave us? Are we any close to defining the network fabric than we are at defining the cloud?
Not really, but that's not necessarily a bad thing. A hard and fast definition would introduce too much rigidity on a technology that by definition must remain flexible in the face of rapidly changing data environments.
While virtualization is the underlying technology that enables the cloud, fabric technology is its most crucial component, but it won't come easy and it won't be one-size-fits-all.
Arthur Cole covers networking and the data center for IT Business Edge. He has served as editor of numerous publications covering everything from audio/video production and distribution, multimedia and the Internet to video gaming.