Is it me, or does the network industry remind you of Revenge of the Nerds
? Networking was cast aside in the cloud revolution, but now companies are learning - the painful way - what a mistake that was. Don’t kid yourself one bit if you think that VMware’s acquisition of Nicira
was mostly about developing heterogeneous hypervisor data centers or reducing networking hardware costs. If you do think that, you’re probably an application developer, hypervisor administrator, or data center architect. You’ve been strutting your newly virtualised self through rows of server racks over the last five year, casually brushing aside the networking administrators. You definitely had some outside support for your views: Google, VMware, and even OpenFlow
communities have messaged that networking organisations aren’t cool anymore and need to be circumvented by coding around the network, making it a Layer 2 network or taking over the control plane.
To be fair, though, networking vendors and networking teams helped to create this friction, too, since they built their networks on:
- 40 years of outdated networking reliability principles. The current state of networking can be in many ways traced back to ARPANET’s principle: a single method to reliably communicate a host of multiple sets of flows, traffic, and workloads. Basically, voice, video, and all applications traverse the same rigid and static set of links that only change when a failure occurs. The package didn’t matter.
- Operations driven by CLI, scripts, and wizards. A typical enterprise network engineer uses CLI to set up each switches’ configuration and all the associated policies. Configuration is complex and requires the devices to be rebooted when a change is applied. HP reports that it takes on average 250,000 command line entries to set up a data center network - making the security, reliability, and performance all fragile.
- Kerberos controlling the network. The network connects users, applications, data, and compute together - but network teams, that have carved their bones on arcane protocols and knowledge, have become the masters of complexity and control the speed and direction of the infrastructure and operations team. Simple adds, moves, and changes to services require the manual manipulation of the underlining components by network administrators.
- Networks run by ticket systems, email, and phone calls. Forrester has found that a majority of organisations use ticket systems, followed by phone calls, and then emails to communicate adds, moves, and changes to the network team. This manually driven way of communicating infrastructure change drags out response time.
Yesterday’s announcement means that networking personnel can come out of the basement, blink a few times and say, “ I'm a (networking) nerd, and I'm here tonight to stand up for the rights of other (networking) nerds. I mean uh, all our lives we’ve been laughed at and made to feel inferior
those bastards, they trashed our house. Why? ‘Cause we’re smart? ‘Cause we look different? Well, we’re not. I'm a (networking) nerd, and uh, I'm pretty proud of it.” Yes, that’s from “Revenge of the Nerds.”
Now that the industry acknowledges that networking is a vital component of infrastructures, we now must face a deluge of questions: Does networking move out of the hands of networking personnel to other teams per VMware’s and Google’s vision? Or is networking an integral part of hardware and software solutions we see being developed within smokestacks at HP, IBM, Dell, and Cisco? Where do Layer 4 to Layer 7 solutions — firewalls, application delivery controllers, WAN optimizers, cloud gateways, etc. — play in this new world? Can we really have software-defined data centers without including those services? Will the network administrator’s job go the way of the Detroit autoworker? Who should you hire to deploy and manage this new world? Is it a person with networking or software background?