While the software-defined data center offers benefits such as increased agility and efficiency, few enterprises have realized those benefits. Limitations with system management tools and IT skills have especially stymied adoption.
Virtualization is the key enabler of the software-defined data center (SDDC). Rather than tie software closely to its underlying hardware, an abstraction layer exists between the two. This layer enables data center administrators to connect, configure, troubleshoot and manage computing resources in more efficient and inexpensive ways.
A closer look at SDDC architecture benefits
A big benefit of SDDCs is that they help break down the traditional IT silos between compute, network and storage resources. An SDDC automates the delivery of IT resources through a unified process rather than through resource silos. This means, in theory, instead of using multiple tools to manage a variety of technologies, an SDDC architecture provides administrators with a single platform to deploy, monitor, update and scale server, storage and networking resources.
In addition, the abstraction layer masks hardware idiosyncrasies, so organizations no longer have to pay a premium for their primary vendors' technologies. They can deploy low-cost, commodity hardware rather than expensive and proprietary systems.
Finally, SDDCs promise new levels of automation and orchestration. The software foundation enables admins to automate mundane configuration, maintenance and management tasks. With an SDDC architecture, IT teams can accomplish previously complex, time-consuming IT processes, such as adding more storage capacity, in fewer and simpler steps.
Hitting a few bumps
Torsten Volkmanaging research director, Enterprise Management Associates
The path to SDDC architecture, however, has been strewn with potholes, both technical and managerial. This is, in part, because components in an SDDC adoption have evolved at different rates. Server virtualization is now widely deployed; in fact, many organizations have virtualized more than 75% of their x86 server workloads, according to Gartner. Using tools like VMware and Microsoft's Hyper-V, admins are accustomed to whipping up new virtual machines with a few keystrokes.
Storage and network automation, however, have been slower to take hold. Storage management involves physical disks and software infrastructure. Currently, those two components are still closely tied together, so changes require manual intervention. In addition, many organizations race to keep up with skyrocketing data growth; market research firm IDC expects data volumes to double every two years and reach 40 zetabytes in 2020. Data center staff often does not have the time or ability to transform its storage architectures to meet these demands.
Similar problems are evident in networking. With software-defined networks (SDN), overlays separate physical and virtual services. The physical network infrastructure typically serves as the underlay network. Theoretically, admins can instantiate, extend and remove virtual overlays dynamically, but implementation has not been simple.
One reason for this is that SDN vendors rely on different protocols and application programming interfaces.
"Tying a Brocade and Cisco switch together is not easy," said Torsten Volk, managing research director at Enterprise Management Associates, an IT analyst and consulting firm in Portsmouth, NH. Open source standards could help IT organizations reduce vendor lock-in, but no universal management standards have emerged.
Dealing with this mixed bag
The end result is an SDDC architecture that features a hodgepodge of different vendor management tools. While businesses can work to tie these various elements together, that process, in reality, has proven difficult. And, without a common infrastructure to build on, only barebones orchestration layers have been available to date.
"The problem is the management service layer is largely missing, and businesses do not want to write it themselves," Volk said.
In some cases, system monitoring and troubleshooting have become more complex with SDDC adoption. In traditional data centers, traffic flowed in a consistent manner, so management tools could collect performance information and identify aberrations over time. Now, with the dynamic nature of system overlays, management tools lose the ability to correlate traffic patterns. In addition, traditional tools encounter blind spots and inconsistencies; virtual connections move traffic in different ways at different times, so recognizing a problem becomes more difficult without a common baseline.
Consequently, enterprises have been forced to keep their traditional management tool sets and use them to pinpoint performance bottlenecks. Often, these legacy systems have limited hooks to new management tools. Rather than the promised "single pane of glass," admins bounce between a variety of autonomous systems to pinpoint problems.
Smoothing out the transition
Providers are taking steps to smooth out the rough edges. Top infrastructure vendors, like Cisco, Dell EMC, Hewlett Packard Enterprise and VMware, have been extending their management systems to new virtualized, software-defined environments. Established management platforms from vendors such as BMC Software, CA Technologies and Microsoft have also pushed into this new space.
Human capital represents another challenge. As traditional IT management activities become automated, the role of the data center staff changes. With an SDDC architecture, these individuals no longer tinker with system innards; instead, they create, move and manage workloads dynamically. Consequently, they need to think strategically and not technically.
"Many IT staff are more comfortable working in system silos than with integrated solutions," said Dave Russell, vice president and distinguished analyst, storage technologies and strategies at analyst firm Gartner. Companies need to guide their staff through this transition period.
The end result is the SDDC market has stalled. "The issues we are talking about today are the same ones we were discussing three years ago," Volk said. In the long term, vendors are expected to put more building blocks in place to ease the transition from old to new, but, in the short term, barriers are expected to continue slowing deployments.
SDDC lessons can be hard to learn
How to plan for an SDDC architecture
Evolve management practices for SD-WAN