Earlier this month I attended the Gartner Data Center Summit, and sat in on several presentations, including David Cappuccio’s intro keynote on data center trends and Neal MacDonald’s session on virtualization security. I apologize in advance for the breadth of my rambling and speculation, but I think the IT revolution playing itself out will mark the evolution of virtualization into the network and into private clouds built within a new type of elastic data center. While I cite two excellent Gartner presos the following thoughts are my own extrapolations.
Surprisingly, there was a common theme between these two otherwise different (Cappuccio and MacDonald) presos catering to two distinct IT audiences: Cappuccio mentioned “vertical scalability” several times while MacDonald talked about enterprises increasingly embracing virtualization for agility (versus the initial driver of efficiency-driven economy). My imagination was stirred by their insight.
Data Centers Designed for Growth Instead of Obsolescence
Vertical scalability is the ability for a data center to increase power density as needed, without major modifications. It is the room to grow within a specified space, with minimal disruption to IT operations. You could consider vertical scalability a form of business agility; as your data center needs increase you can grow without having to build a second building. That can mean a massive capex savings. If you are in a highly efficient (PUE below 1.30) vertically scalable data center the payoff accelerates with utilization, accelerates with growth. Compare that to the traditional data center model where growth leads to increasing capex and opex.
Network Automation and Consolidation
If we take the agility concept a step further into the bowels of IT one cannot help wonder what will take place with the myriad of network hardware appliances sitting in path between increasingly agile application and storage infrastructures. There is precedent for consolidation and automation in the network. Remember the compression or caching appliances that once sat in line with server load balancers? We used to call that world packet ping pong. That was the world of networking before the consolidation that ultimately led to the application front end or (what Gartner calls) the application delivery controller.
But why stop consolidation at layers 4-7 in the OSI stack, now that cloud operating models are driving IT teams to be more elastic, more responsive? In the same way that virtualization tore through the legacy one app one server world and created new agility potentials, what is to stop software from marching further through the network appliance world to make networks similarly easier to manage and more elastic?
Software and the Network
Marc Andreessen said as much in the Wall Street Journal on Why Software is Eating the World, albeit his focus was on consumer technology. And if you weave together Cappuccio and MacDonald messages it seems the vision is clear: IT either becomes the quick or the (replaced by the cloud) dead; and that is ultimately a software-centric challenge. As server admins discovered in devtest, the abstraction layer that untethered the application and workload from specialized hardware generated impressive efficiencies of scale.
The Data Center as a New Point of Strategic Advantage
That leaves the legacy network and the data center on notice. The manually managed, legacy appliance network and the traditional (walls and generators) data center begin outliving their usefulness as IT becomes more elastic. Networks will need to become more dynamic, more people–efficient and data centers will need to become more energy efficient, more scalable and more custom to the unique demands of a more elastic IT organization.
The data center becomes the strategic point of leverage as the IT infrastructure becomes increasingly elastic. Legacy data centers become obsolete at a faster pace, simply because they confine IT at a time when IT needs more agility and efficiency.
A single next generation data center (vertically scalable and highly efficient at a wide range of utilization levels) can give a team unprecedented agility while reducing power and cooling costs by millions of dollars per year. Elasticity and economy: the rallying cry of virtualization, networks and the data center. There I said it.
Will the Network go x86?
When Neal was sharing his slides on virtualization security and talking about the tech and business case migration issues (or the difficulties of network security transforming into virtualization security) of various vendors aimed at addressing the transforming data center, the obvious network agility solution seemed to lead to the x86 network: one off appliances replaced by inline workloads managed by mouse clicks supported with unprecedented automation across a backplane of increasingly powerful network servers powered by the latest commodity processor.
That is indeed a tall order for many vendors with specialized chip architectures. Oops.
The Virtual Tether
The legacy network and the link between the physical appliance and network application was the cornerstone of network security. Yet to unleash business agility that physical coupling needs to be replaced by a kind of virtual coupling, which makes the otherwise mundane world of IPAM strategically interesting to the necessary evolution of the network and virtualization, from management to security.
Yet as IT evolves, you can expect more pressure on data center design and construction in the form of demands for vertical scalability and high energy efficiency, especially from customers consuming more than 500 kW. At 1MW and higher IT teams should expect more input and control over electrical and mechanical architectures, versus one size fits all (take it or leave it) designs. Customization and efficiency can drive seven figure annual returns in elasticity and efficiency.
Yes, commoditization inside the data center drives added-value for specialization and customization of the data center itself, based on the unique needs (applications, growth rates, latency, availability, power consumption, local climate, etc.) of an IT organization under increasing pressure to do more with less… at an even faster pace. One can either containerize at the low end or customize at the high end, or get trapped into commodity space that becomes more painful as your needs grow, just like the legacy network and yesterday’s legacy server farm.