Lately, VMware has put a focus on what it calls the software-defined data center, but the larger idea of virtualizing the infrastructure and delivering it as a service isn't specific to VMware.
Rob McShinsky, Dartmouth Hitchcock Medical Center
Software-defined data centers are a reality today. Virtual switches (like Cisco's Nexus 1000V), dynamically optimizing storage, physical server reassignment, automatic VM load and power balancing are all major data center components that are completely reliant on complex software.
The problem is, the software that controls, provisions, monitors and allows this puppetry are all islands of their own or components of converged infrastructure with a focus on a small subset of hardware and software vendors.
A true software-defined data center would allow for more generalization of every component, as long as standard interfaces to the varying components were enacted. The mixing of storage, server, network and hypervisor providers, all controlled by a software director, is a novel idea. We are starting to see the future of data centers emerge in the hypervisor management space with VMware's recent foray into the multi-hypervisor market with vCenter Multi-Hypervisor Manager and Microsoft's System Center Virtual Machine Manager. Both vendors are attempting to completely abstract the hypervisor and, to a degree, many hardware components as well. Some may still disagree, but the hypervisor itself has been greatly generalized, and the competitive battleground is now this breadth of management software.
A more broad-reaching software product would need to rely on strict standards. With profits tied to differentiating vendor capabilities, it is difficult to see this happening in the near future. For now, look for converged infrastructure alliances to broaden to fewer players with more hardware vendors buying management software assets.
Dave Sobel, Evolve Technologies
VMware may call it the software-defined data center, but it's really delivering on the promise of cloud computing and virtualization. As cloud computing becomes more and more a part of how we approach using technology effectively, businesses will demand that computing resources be available on demand and as needed. This is only achievable through software, as any dependency on hardware makes true on-demand computing impossible. To me, the future of data centers and the idea of the software-defined data center is the ability to add resources via self-service, exposing all the possible configuration options as needed, to be generated completely on demand. This is easily achievable for some components now -- CPU, memory and disk resources are all easy to deploy. However, that's not the case for many other components, like network connectivity, multiple Internet service provider linking and cross-service failover.
I look forward to a world where computing technologies move from data center to data center, from provider to provider, all on demand and all through software. Many of the challenges will be business- and process-oriented, and we must strive to build common language and standards to allow for this. It's critical as data becomes larger that this flexibility be built, allowing for movement from cloud to cloud.
I do think we will get there; the question is simply how long it will take. The economics of moving to the cloud make such sense, and the bandwidth issue will be solved over time. Ultimately, the economics will work in favor of the service provider as well.
This was first published in March 2013