It’s only natural that talk of software-defined networking (SDN) should lead to talk of the software-defined data center (SDDC). But since large portions of the data environment (everything from the virtualization layer and up) are already defined by software, the real change taking place is more accurately described as software-defined infrastructure (SDI).
When we think about the slow pace and overall hassle of provisioning and repurposing application environments, it almost always comes down to a hardware problem. Not only is hardware a finite commodity in the data center, but it requires a lot of coordination and manual coding to ensure that resources can communicate and interact properly. In today’s “gotta have it now” business environment, these kinds of artificial roadblocks just won’t do.
In fact, when you consider all the promises of resource flexibility and broad federation that were made when both virtualization and the cloud were brand new, they were actually based on the idea of SDI, says Convirture CEO Arsalan Farooq. Now that data center managers, and even users, can easily spin up applications and architectures on a whim, SDI is the only way to prevent this complexity from overwhelming both the physical and virtual resource sets. Under SDI, the correct technology decisions can be made easily, or undone and then revamped should initial assumptions prove wrong, leading to an enterprise that is optimized for applications and users rather than restricted by available hardware.
This capability will emerge as a crucial component to the mega data centers that are already emerging in the cloud era, says Wikibon CTO David Floyer. Economic pressure will cause small and medium-sized enterprises, and perhaps even a few large ones, to turn to giant regional facilities with high numbers of tenants all striving for available resources. Traditional provisioning methods would be a nightmare for such an environment, so SDI will be the only way to put resources into play on a timely basis and ensure that utilization and other factors are kept in check to prevent operating costs from getting out of hand.
SDI is also one of the key aspects of new virtual and cloud platforms, such as Cisco’s Global Intercloud. The company had long resisted the urge to branch into services that would compete against Google and Amazon, but at the end of the day an integrated application and services suite tied to software-defined infrastructure was deemed too lucrative a value proposition to pass up. In one move, Cisco can now present its entire portfolio – from the Application Centric Infrastructure (ACI) to the Unified Compute platform to any number of smaller offerings dealing with Big Data or virtual desktops – as an integrated end-to-end, top-to-bottom virtual computing environment for organizations that prefer to devote resources to building markets rather than building data infrastructure.
The enterprise should beware of any hardware company proposing software-defined anything, says VMware’s The SANMAN. At best, you’ll most likely see traditional hardware infrastructure rebranded as SDI, but at worst you’ll get a version of SDI that does not extend itself across multivendor commodity hardware as well as it should. The beauty of a pure SDI play is to allow users to define the environment that best suits application requirements. In that vein, is it wise to implement SDI control software developed by one of the vendors of your legacy hardware? Of course, VMware has no stake in hardware, but it is the chief developer of legacy virtual environments, so it is fair to ask whether it is worth federating the hardware layer by adopting a single-vendor virtual and SDI environment.
Ideally, open protocols like OpenFlow and OpenStack would produce a vendor-neutral resource ecosystem, but it is hard to tell how this is shaping up at this stage of the game, given that nearly all the top enterprise vendors are working feverishly to leverage these formats for their own platforms.
At the end of the day, infrastructure providers will have to rely on knowledgeable minds in the vendor community to guide them through the transition from physical to virtual to cloud to SDI. Small wonder, then, that many organizations will simply come to the conclusion that it is easier, and cheaper, to lease data resources than build them.