While working to support a large-scale Enterprise cloud migration program, I’ve been giving a lot of thought to the question about what it means to be an Enterprise IT organization at the far end of such a migration. There are many wide-ranging impacts, of course, but one of the near-term considerations is what it means to “operate” in the cloud.
In the world of data centers and large infrastructure teams, one of the most fundamental principles is that efficient operations require a set of broad tools which can be used to control a variety of common functions. This led to the growth of large suites of IT Operations and Management tools from IBM, CA, HP, and BMC in particular. These tools allowed systems and operations teams to centrally maintain, monitor, and control the vast array of servers the Enterprise IT teams were responsible for in a scalable manner. And it made a lot of sense. Without this type of control plane it was almost impossible to maintain any semblance of cost or operational control, and I’m certain that many implementations were driven by embarrassing outages, expensive security breaches, or aggressive cost control measures. “We’re a big IT shop. We need a strong mechanism for control.”
Enter the cloud. Most early adopters of Infrastructure cloud services in the Enterprise (running almost entirely on Amazon Web Services) were isolated, and often outside of the IT department itself. The notions of operations and control were either foreign or the frustrating bottleneck to escape. But as The Cloud started to attract executive attention, either because of concerns or excitement about these fringe internal activities, several things started to happen. The first was the turn of attention from existing Enterprise IT vendors, who started to develop their own cloud-like services, typically built as software solutions to be deployed on their existing lines of hardware. Next, seeing the increasing momentum of public cloud services, they began propagating the idea of the hybrid cloud, the pleasant notion of simultaneously running operations in both on-premise “private” clouds, as well as in public clouds like AWS, Microsoft Azure, and Google Compute Engine. Simultaneously, a few start-up vendors emerged–in particular Rightscale–who were looking to make multi-cloud deployment and management possible. While initially this market focused on public cloud environments, the natural evolution, validated by the entrance of the traditional vendors as well, was to look at the full portfolio of enterprise cloud operating environments, and so emerged the Cloud Management Platform market.
With these vendors came a commonly repeated product aspiration: to provide Enterprise IT with the “single pane of glass.” This would allow Operations staff to keep track of all of these new cloud things, move workloads seamlessly between private and public clouds, and do all of those Very Important Activities that were critical for central IT to perform. It sounded great! It made perfect sense to today’s IT executives, and so was an easy sell for vendors–new and legacy–who wanted to keep up the sales pipeline. The same Enterprise Control Plane, re-imagined for the cloud world.
The “single pane of glass” IT operations model: Rational. Sensible. And completely wrong.
The problem with the application of the Enterprise Control Plane to the cloud is that it brings forward all of the old assumptions about what is necessary to securely, effectively, and efficiently manage infrastructure. When Infrastructure is a tangible thing, you first need a way to manage all of its physicality–power, cooling, installation and maintenance of the hardware and its components. You had to know if something was going wrong, and know when to get it fixed, and you replied on your operations software to do this. Hardware in the cloud? It’s still there, but you don’t have to care any more.
Next, the actual systems need to be built and installed. While this has become less of a craft business than it once was, building the necessary components to make it more scalable–leveraging, of course, your newly licensed add-on component for your operations software–is still labor and capital intensive. While the cloud doesn’t always eliminate your specialized system designs, in many cases it can, and at the very least you don’t need to buy, install, and configure all of the extra software to make it simple for your community–whether they be developers or system admins–to deploy new servers.
Within operations, a large control plane was how ongoing functions such as releases, configuration management, monitoring, logging, and auditing could be performed, and these were the domain of Infrastructure professionals. Again, these are all still necessary in the cloud, but an entirely new set of services and tool chains have democratized their implementation and use.
The net of all of this is that the value once provided by infrastructure staff–and the hardware and software they managed–has been simultaneously commoditized and dispersed. And as the data centers begin to close, the era of vast Enterprise IT control planes will close as well.
The functions once provided by the Enterprise Control Plane have not become irrelevant or unimportant–it’s just that the locality and means by which they are provided has changed. A large part of their one-time value is now embedded within the services offered by cloud vendors, and another segment can be adequately managed closer to the point of value, namely within the business or software management and development teams. There will likely remain services that look a bit like Infrastructure today, but the emphasis will shift from one of control to one of enablement, a means to support the actual business objectives more directly and in a far less monolithic fashion.
While it won’t collapse, I think it fair to say we are at the end of the era of the Enterprise Control Plane, and are at the start of the era of Enterprise IT Enablement. This isn’t a value judgement, only the recognition that the IT Infrastructure service model which made sense for so long is radically changing, and with it so must our assumptions about what is required to deliver and support services in this new world.