Software-defined data center: a strategic journey, not a technical destination

brian blog
Dell EMC

Software-defined data center: a strategic journey, not a technical destination.

Your car might be partially software-defined without your knowledge, just as I found out not long ago.

Now I don’t drive a Tesla, which was built from the ground-up with a majority of software-defined components. I drive your everyday, traditionally designed vehicle. Just by looking at it, you’d have no idea it included software-defined components. So how did I stumble upon this fact?

I took my vehicle in for a service not too long ago for something that we can all agree is a critical issue – my radio and speakerphone weren’t working properly. At the shop, the service advisor and the vehicle engineer took the car for about 30 minutes to “gather data” and then returned with some very specific advice: “Next time this happens,” they said, “we want you to turn on the hazard lights or tap the brakes five times in a row.”

At first I thought they were trying to prank me. But, after a quick brain recharge with some espresso, I realized something quite interesting – they were asking for a beacon in the logs to help them find the error.

When I brought the vehicle back in, they admitted the patient for 24 hours. During its stay, they contacted the radio manufacturer, had a patch developed and deployed that patch onto the vehicle, completely resolving the issue. My vehicle, it turns out, is partially software-defined. Cool.

Software-defined implementations might be subtle, like in my car; they might be incorporated from the ground-up, like in a Tesla – there’s no one right or wrong approach. Accordingly, your individual path to the software-defined data center is not the same as the company on the floor below you. Nor is it the same as your biggest competitor. It’s your journey based on your specific needs, current capabilities and future investments.

Tesla set out to build a car that would deliver a combination of luxury and performance while bringing the electric vehicle into the mainstream. But other manufacturers already have vehicles, roadmaps, designs, etc. and need to be able to embrace technology at a completely different pace and methodology. If Tesla or some other company set out to make a fully software-defined vehicle from scratch today, it probably would not be another incredibly stylish car with a 17” screen in the console. Instead, it might look like a modern living room or conference room with wheels. The same can be said for your software-defined data center. You do not have to start from square one to make an impact with software-defined technologies. It’s more important that you evaluate each project outcome and then decide if software-defined would help you better achieve your goal.

Software-Defined Data Center (SDDC / SDX) The fundamental concept behind a software-defined data center is to abstract individual physical resources (such as compute, storage and networking), pool them together and then virtually present them, either as a whole, or in parts, to guest operating systems or workloads for easy consumption. While software-defined technologies are not new, there are some major forces at work accelerating their adoption. Moore’s Law, and the opportunities created as processor speeds consistently double, enabled software-defined pioneers to deliver enhanced services on top of new, more powerful engines. As a result, specialized hardware is increasingly being replaced by specialized software in your technology landscape.

Software-Defined Compute (SDC) Today, it’s relatively difficult to find compute environments that are not software-defined and running on some form of virtualization platform. Between VMware, Hyper-V, KVM, Docker, Mesos, Kubernetes and the ever-growing list of virtual machine/container platforms, a vast majority of compute workloads are running abstracted from the physical compute beneath them. One of the clear benefits of Moore’s law is the ability of those applications to share space on the same processors as the host operating system. Only by exception are workloads not virtualized - for either support reasons (from risk-adverse or resource-constrained third party vendors) or for performance reasons (where resources are directly assigned 1:1 with the hardware they run on). The default assumption has flipped in the case of software-defined compute in the data center today: the choice is whether NOT to virtualize or containerize the workload, rather than whether to virtualize.

Software-Defined Storage (SDS) One of the most heavily discussed sectors in the software-defined data center infrastructure today is software-defined storage. Long gone are the days of looking at software-defined storage for non-mission-critical workloads, archive tiers and other low priority use cases. Why? It’s a perfect storm of technological opportunity. We are now combining the existing benefit of available CPU overhead with the additional impact of flash media performance. Flash costs are also on the decline, it is currently similar in price to 15K drives and shifting quickly to approach the price of 10K drives.

Now that it’s priced so accessibly, how do you harness the speed and response time that flash offers? By moving that flash media from a pooled environment on a fabric and placing it back into the powerful standardized x86 servers, where it’s closest to the CPU. Shortening the communication distance between the two produces phenomenal performance and response time gains with very little trade-off in compute performance. Combine that with the proliferation of 10Gb network in servers and top-of-rack, and flash can deliver performance, as well as data consistency, at scale reliably. As a result, it is clear that software-defined storage is ready for a majority of your workloads by volume today. Software-defined storage is the next logical step that organizations need take in their software-defined journey and it is often done by leveraging hyper-converged deployment models.

Software Defined Networking (SDN) As previously mentioned, with the adoption of software-defined architectures, the overall considerations of the network and its impact in the modern data center have significantly increased. The network plays a critical role in each software-defined function’s ability to communicate between nodes while leveraging scale-out standardized x86 servers. Furthermore, the ability to abstract and programmatically manage network objects (such as interfaces and policies) is the next wave of software-defined adoption, as organizations look to transform their IT operations and gain efficiencies. Software-defined networking allows you to centrally manage those resources and gain a new level of agility. It allows for unique architectures, like microsegmentation of virtual machines or containers. As your organization works to adopt its multi-cloud strategy, understanding your ability to leverage investments in skillsets and technologies as they relate to software-defined networking can derive huge benefit. Software-defined networking is solving problems and delivering efficiencies in securing, shaping and managing traffic that traditional networking simply could not. You don’t have to, but by leveraging software-defined networking you are simplifying your organization’s ability to execute on a multi-cloud strategy.

There is no singular way to take your software-defined journey In case you haven’t already heard, software is eating the world. Like the car manufacturer, you may already have existing investments in place - that’s not a problem. At the next opportunity to evaluate a project, consider how software-defined data center technologies might impact your outcomes. That solution could very well be a small bespoke deployment for a single project or it could be a complete overhaul of a technology or solution set (such as replacing a traditional SAN with a software-defined one). You might also consider employing a more aggressive strategy in which all new projects are evaluated as software-defined first. This approach will shift your overall focus to determining which workloads still require a traditional infrastructure. Whether you’re deploying it side-by-side with your existing infrastructure or going all in on the software-defined model for your entire environment, the technology to support you is more than ready.


Brian Carpenter (@intheDC) is a Director of Marketing for Dell EMC focused on Converged & Software-Defined strategy. He’s also a co-host of The Hot Aisle podcast. When he’s not talking to customers about breaking down silos and thinking a little differently, he likes to take pictures of his food. That is, after he separates the food on his plate to make sure the peas don’t touch the mashed potatoes.

Copyright © 2017 IDG Communications, Inc.