Layer 2 Rant

Before I begin this rant I want to make it very clear… I am a campus network engineer by trade and experience, and now I’m tasked with running operations in higher ed. I don’t touch carrier, I don’t touch MPLS, if my BGP config were any more basic it would be static routing, I don’t have a huge sprawling datacenter. I’m just a guy who tries to get my job done everyday and let my users work. I wear quite a few hats, but I am no specialist in anything, not an expert, not a guru. Now that that’s out of the way…


During the VMworld keynote this year, there was a lot of talk about networking; how to enhance it within the VM environment, how to maximize capabilities, and how to extend it across geographic locations. It’s clear based on what we’ve heard this year, and the recent Nicira acquisition that VMware is ready to not just play nice with the networking folks, but finally stepping up their game and starting to innovate. Hurrah! Right? As an operations manager I am tasked with running everything; VMs, physical hosts, networking, and client facing applications all fall under me. On top of that, as the leader of operations I have to know it all, jack of all trades, good at everything expert at nothing. Due to our staff size and how varied our jobs are I agree completely. Simplified server mobility from one datacenter to another (or into the cloud) makes all of our jobs easier. Running critical applications from redundant locations for load balancing/high availability and migrating between them is nothing more than a few clicks. Disaster recovery/business continuity situations become more like unscheduled migrations. It’s all very enticing.

There has been development for years and innovators are constantly pumping out new methods and protocols to accomplish these goals. The root of the problem isn’t fixed though. There is a fundamental underlying issue here that isn’t being addressed. The network is picking up where the hypervisors and OSes themselves are falling short. Shouldn’t mobility involve all parties? It’s clear the networking industry is capable of fixing the problem and is highly committed to it. Now it’s also (almost) clear that VMware is buying in, but still at the layer 2 level. We’ve been given VCE by these large companies to accomplish a convergence, an ease of deployment, something different. Yet it has failed to deliver something I think it truly missing. It’s 2012 now, virtualization is nothing new. The need to move systems from one location to another, be it across a room or across the globe, has existed almost as long. Why aren’t there any protocols that have been developed allowing the OS to actually understand where it is? Granted, it would take a lot of work to get the major players in a room to discuss and develop it. If not that (and this could be where VMware makes a play with Nicira and Openflow) let the hypervisor become completely aware of network and VM, acting as the mediator in this crazy game of mobility.

Complexity keeps being added in this downward spiral that remove the ability for smaller shops to utilize the existing methods. Whether it’s time, salary, or equipment, implementing mobility right now is expensive.

Maybe someone is working on it. Maybe it’s been said before. Maybe I’m insane. Maybe I’m just really tired and getting punchy from all the server guys taking jabs at the network and me having to bite my tongue. If I’m missing something, if I’m flat out wrong, by all means, tell me so I can get educated.


Leave a Reply

Your email address will not be published. Required fields are marked *