Closing In on the On-Demand Data Center
Technology serves as a means to a business end. In the case of SDN, the end is the on-demand data center. What does that mean, and how are vendors working to get there?
It’s long been said that enterprise infrastructure development should focus on solutions, rather than technology. All the systems, gadgets, software and platforms that comprise today’s data center are simply a means to an end. If enterprise executives fail clarify that end, then all the capital devoted to building and maintaining their infrastructure serves little or no purpose.
On the networking level, the latest technology is SDN. But have we really thought through to the culmination of a truly virtual, fully federated, automated infrastructure? In other words, we're developing the means, but do we know to what ends?
For some, the ends are the litany of advantages that virtual and cloud technologies promise to bring: lower costs, greater scalability, self-service provisioning, etc. These are all worthy goals, but they nonetheless represent a patchwork view of data operations. Once you’ve put all those things together, it becomes clear that the end game in all this is nothing less than the on-demand data center.
The concept is simple enough on the surface but rife with difficulties when applying the mechanics to real-world production environments. It requires not only the most advanced networking systems on the market, but state-of-the-art automation tools capable of giving users what they want in a highly dynamic setting. It’s not clear whether the IT industry can actually produce such an environment at scale. The conversation has reached the point where we can concede that it is possible, though.
Brocade, for one, has adopted the on-demand concept for its new virtual switch portfolio, intended to provide an end-to-end virtual fabric capable of pushing multitenancy into the millions. This would give virtually all users on the system their own virtual networks, to be provisioned and customized as they see fit. With the new Vyatta 5600 vRouter in place, enterprises will not only be able to separate the control and data planes as in typical SDN deployments, but then scale the data plane linearly to multiple cores on the network even while the control plane is safely housed within just one or two cores.
Juniper is on a similar track with its NetworksContrail system, which combines standards-based network virtualization (CloudStack and OpenStack, but still no OpenFlow, apparently) with network intelligence tools to embed advanced networking capabilities into the hypervisor in support of broad automation and orchestration across multiple cloud platforms. Through techniques like automated provisioning and dynamic service chaining, the company intends to enable on-demand network services and infrastructure across standard x86 environments. IBM has already agreed to integrate the Contrail system into its SmartCloud Orchestrator. Juniper is also releasing an open source version of the software, dubbed OpenContrail.
How would this level of functionality change things on the ground? We can already see some of the basic elements taking shape in the financial industry. Fidelity Investments recently teamed up with Integrated Design Group to build an off-site “fabricated” data center capable of scaling load capacity on-demand to help the company more closely match resource utilization with data requirements. Part modular infrastructure, part kit-construction, the Centercore system is designed to deliver a scalable solution that meets the requirements of high-volume financial trading environments, with rapid deployment and low upfront capital costs. The design has pulled in a number of honors, including the 2013 CoreNet Global Industry Excellence Award and DatacenterDynamics’ top choice for “Innovation in the Medium Data Center.”
Technology, and product development in general, has always been about giving people what they want, quite often before they even realize they want it. In the case of the on-demand data center, though, it isn’t hard to envision all sorts of productivity and competitive advantages coming from allowing users to provision the data resources they need at a moment’s notice. It seems, then, that we finally know what we want, and we are tantalizingly close to getting it.