Over the last five years I’ve watched server virtualization become ubiquitous in most data centers, based on the promise of higher infrastructure and increased business agility. I’ve also watched server administrators throw up their hands in frustration, as network administrators have been unable to provide the networks needed to harness the true value of the virtualized infrastructure. Server admins want to drop VMs into a big pool of compute capacity, not limited to a select number of servers in a rack or row, and networks just aren’t that flexible.
The good news for server administrators is that the concept of overlay networking enables just that. Overlay networking delivers the ability to move a VM anywhere in a data center or even between data centers by making a broadcast domain a ubiquitous service available anywhere in the network. The bad news for network administrators is that the server teams are building and operating it themselves.
As a network engineer, I can’t really blame them. I’d like to think that we build networks for the sake of building networks, but the reality is we build networks to connect users with applications. The server administrators are our customers and the service we are providing them is antiquated and no longer meets their needs. Like any other business, if you can’t deliver what your customers are asking for they will go elsewhere.
If it’s not clear yet I’ll state it more bluntly: Our networks are antiquated and we’re on our way to becoming a much smaller part within the organization. We are the next mainframe operator in the IT organization; the grumpy guy in the corner nobody wants to talk to (sorry mainframe guys). We use manual (read slow) provisioning models, our equipment is relatively expensive, and our networks don’t provide the flexible connectivity our customers need. You only need to look at products like the VMware vSwitch, HPs Virtual Connect, and Cisco’s UCS to know this is true. These products we designed to bypass network administrators and the ‘terrible’ service we provide.
Take what little comfort there is when I say this is not our fault. We didn’t architect our networks poorly or forget to turn on some obscure feature in our switches or routers. As a group we are doing exactly what we have been trained to do for the last 20 years. The simple reality is that what we did 20 years ago doesn’t work anymore and the entire network industry is to blame for not adapting. Cisco, long been the leader in the space, has been slow to drive real innovation in my opinion. The rest of the industry is still stuck in a business model where being a cheaper version of Cisco will get you a couple points of market share which simply compounds the problem.
So what can network administrators do about it? As I see it there are two options. We can bury our collective heads in the sand until it’s all over and the network belongs to the server team or we can start to take a very critical look at the needs of our customers and start thinking differently about networking. I truly believe the best path forward is the later, which is why I find Software Defined Networking so interesting (and the reason I joined Big Switch Networks in the first place). Incase you disagree I suggest you hedge your bet by taking some VMware courses.