When I started my career in IT, infrastructure provisioning involved a lot of manual labor. I installed the hardware, installed the operating systems, connected the terminals, and loaded the software and data, to create a single stack to support a specific application. It was common to have one person who carried out all of these tasks on a single system with very few systems in an Enterprise.
Now let’s fast forward to the present. In today’s world, thanks to the dynamics of Moore’s Law and the falling cost of compute, storage, and networking, enterprises now have hundreds of applications that support the business. Infrastructure and applications are typically provisioned by teams of domain specialists—networking admins, system admins, storage admins, and software folks—each of whom puts together a few pieces of a complex technology puzzle to enable the business.
While it works, this approach to infrastructure provisioning has some obvious drawbacks. For starters, it’s labor-intensive with too many hands in order to support, it’s costly in both people and software, and it can be rather slow from start to finish. While the first two are important for TCO, it is the third that I have heard the most about… Just too slow for the pace of business in the era of fast-moving cloud services.
How do you solve this problem? That is what the Software Defined Infrastructure is all about. With SDI, compute, network, and storage resources are deployed as services, potentially reducing deployment times from weeks to minutes. Once services are up and running, hardware is managed as a set of resources, and software has the intelligence to manage the hardware to the advantage of the supported workloads. The SDI environment automatically corrects issues and optimizes performance to ensure you can meet your service levels and security controls that your business demands.
So how do you get to SDI? My current response is that SDI is a destination that sits at the summit for most organizations. At the simplest level, there are two routes to this IT nirvana—a “buy it” high road and a “build-it-yourself” low road. I call the former a high road because it’s the easiest way forward—it’s always easier to go downhill than uphill. The low road has lots of curves and uphill stretches on it to bring you to the higher plateau of SDI. Each of these approaches has its advantages and disadvantages.
The high road, or the buy-the-packaged-solution route, is defined by system architectures that bring together all the components for an SDI into a single deployable unit. Service providers who take you on the high road leverage products like Microsoft Cloud Platform System (CPS) and VMware EVO: RAIL to create standalone platform units with virtualized compute, storage, and networking resources.
On the plus side, the high road offers faster time to market for your SDI environment, a tested and certified solution, and the 24x7 support most enterprises are looking for in a path. These are also the things you can expect in a solution delivered by a single vendor. On the downside, the high road locks you into certain choices in the hardware and software components and forces you to rely on the vendor for system upgrades and technology enhancements, which might happen faster with other solutions, but take place in their timelines. This approach, of course, can be both Opex and Capex heavy, depending on the solution.
The low road, or the build-it-yourself route, gives you the flexibility to design your environment and select your solution components from the portfolio of various hardware, software vendors and open source. You gain the agility and technology choices that come with an environment that is not defined by a single vendor. You can pick your own components and add new technologies on your timelines—not your vendor’s timelines—and probably enjoy lower Capex along the way, although at the expense of more internal technical resources.
Those advantages, of course, come with a price. The low road can be a slower route to SDI, and it can be a drain on your staff resources as you engage in all the heavy lifting that comes with a self-engineered solution set. Also, it is quite possible with the pace of innovations that you see today in this area, that you never really achieve the vision of SDI due to all the new choices. You have to design your solution; procure, install, and configure the hardware and software; and add the platform-as-a-service (PaaS) layer. All of that just gets to a place where you can start using the environment. You still haven’t optimized the system for your targeted workloads.
In practice, most enterprises will take what amounts to a middle road. This hybrid route takes the high road to SDI with various detours onto the low road to meet specific business requirements. For example, an organization might adopt key parts of a packaged solution but then add its own storage or networking components or decide to use containers to implement code faster.
Similarly, most organizations will get to SDI in stepwise manner. That’s to say they will put elements of SDI in place over time—such as storage and network virtualization and IT automation—to gain some of the agility that comes with an SDI strategy. I will look at these concepts in an upcoming post that explores an SDI maturity model.