Hardware abstraction layers are becoming more and more common in our infrastructures: we began years ago with the hypervisors (CPU and RAM), now there is the wave of totally virtualized storage solutions (next generation scale-out storage arrays and VSAs) and the next big deal will be networking virtualization (Nicira and Embrane are only a couple of examples). Are you ready?

Why software… a brief history

The first x86 hypervisor was developed to virtualize a PC into a PC, mostly for development tasks, then it grew to the server side and finally it became the first software layer on top of commodity x86 server.
At the beginning of server virtualiztion the virtualized infrastructures were still in use only for development or tier 2 services but things changed quite fast year after year.
Server side virtualization showed its firsts limits when numbers increased: big infrastructures suffered physical storage management costs, lack of good resources utilization, performance and scalability.
The only way to solve these issues was developing “virtualization friendly storage” through a tight software integration. In fact, the most virtualized/integrated systems, the ones with the best software features, fit at best in deep virtualized infrastructures.
As one problem is solved, another arises immediately: network! Now you can manage tons of VMs on top of a transparent storage layers, you can move VMs (and associated data) back and forth between infrastructures without service disruption. Provisioning, planing and management are easy and the Sysadmin/VM+TB ratio has grown a lot in the last years… all this is fun, at least until you need to touch the network layer.
Right now, Network virtualization capabilities are still very basic (apart some costly hardware stuff), also in the most advanced hypervisors, and every time you need to re-configure your network you are forced to deal with complex management software, cables, firewalls, etc, etc. In big datacenter you can provision a pool of VMs and related storage in a few minutes but you risk to hanging on for weeks while waiting for the underlaying network configuration.

Cloud aggravates the problem

If you think at some of the cloud computing pillars, like automation and thin provisioning, you can get a clear picture about of how a traditional network could be a pain in the a$$.
When you add the cloud layer on top of the virtualized infrastructure the risk is to build a Ferrari with Fiat 500 wheels: probably you’ll go fast anyway but only on very straight roads… power is nothing without control!
The problem is getting even bigger on huge enterprise private clouds or football field sized DataCenters.

Next generation infrastructures

The trail was already pioneered by Servers and Storage, now it’s time for networking stuff. I’ve already mentioned some of the new names in town but I’m sure that others are working on similar solutions and they’ll come out soon… Traditional networking vendors won’t be happy but this is what is going to happen.
A 100% commodity infrastructure, built with swappable standard hardware bricks and topped with a strong virtualization layer, is the goal for many providers and large customers. Eventually, network hardware layer will shrink to super fast stupid layer 2 switches and the all game will be played on a price basis as it is now for servers and soon for storage.

The consequences

During the last HP Discover I read a message somewhere like “in 2020 most enterprises won’t have an IT department”. Well, I’m not sure if your enterprise will kill its IT department (I actually agree that it could be one of the hypothetical scenarios!) but chances are that you won’t even own a physical Datacenter by that date. The reasons to choose a 100% private cloud deployed on a public infrastructure will be the competitive price but also elasticity and features that you couldn’t achieve in your small infrastructure.
With the ability to build real Private clouds (totally virtual DCs), charged with the pay-as-you-go model, many SMBs will easily adopt the IaaS instead of thinking about an internal DC, especially smaller ones. Of course, SMBs will also adopt PaaS as well as SaaS but I’m sure that some legacy stuff will remain on IaaS.

Bottom line

Ordinary infrastructure as we know it will slowly die and, of course, your traditional infrastructure is going to die as well. On the flipside the totally virtualized infrastructure won’t be convenient for the SMBs and they won’t be able to take advantage from the large scale economy of buying tons of stuff, the only way for them will be IaaS (alongside PaaS and SaaS).
It’s also clear to me that while these technologies and IaaS cloud proposition will mature, hardware sales on SMBs will slow down proportionally.
Eventually, an 8 year timeframe is enough to rethink the channel business, don’t you think?