Sometimes called unified computing or dynamic infrastructure, the convergence of computing resources describes “bringing together of compute, storage, and networking in order to facilitate virtual computing and application deployment flexibility,” said Jim Frey, managing research director for networks at Enterprise Management Associates.
Although the details of implementing converged infrastructure can be complex, the basic premise is straightforward. In the old days, businesses divided their computing resources into silos to perform independent or dedicated tasks. For example, a business might own one physical server which behaved as a mail server, and another machine as a storage server, and another as a Windows domain controller. Depending on the size of the organization, each of these may actually have consisted of multiple machines and associated hardware, all dedicated to their assigned task. The problem is siloed computing can be a very inefficient way to allocate resources. Consider:
- Computing power – In a siloed configuration, each CPU may only be utilized a portion of the time. The business is essentially paying to provision, power, and maintain a machine that is only leveraging a fraction of its capacity.
- Storage capacity – Likewise, if the mail server is nearly running out of disk space but there is ample space on the Windows domain controller, this excess capacity is not easily accessed by the task that needs it most.
- Heterogeneous platforms – An organization may need to support software for multiple OSes; dedicating separate hardware to each requires increasing physical assets.
The promise of convergence
At its most extreme expression, converged infrastructure makes a simple proposition: all of your data center assets — computing power, storage space, network connectivity, power and cooling — are one common utility pool from which you draw on for whatever tasks you need at that moment.
In the perfectly converged environment, there are no silos at all. Computing assets are totally elastic, adapting to changing demands on the fly.
Of course, in the real world life is more complicated. Many organizations already have investments in existing infrastructure and cannot justify tossing it all overboard wholesale. “In rare cases,” noted Frey, “there are true Greenfield build-outs where a clean slate exists.”
Still, the opportunity window for converged infrastructure is increasing for three important reasons: depreciation of existing infrastructure, the maturing of virtualization, and the appeal of the cloud.
Driven by virtualization
Converged infrastructure owes its very existence to the rise of virtualization technology. Virtualization essentially rides an abstraction layer between the OS and the underlying hardware. Software running on top of the virtualization layer is agnostic about the physical hardware.
Through virtualization, multiple OS instances can run simultaneously on the same physical machine. These OS instances can be the same or different from one another, allowing one set of hardware assets to host multiple Windows instances, or mixed Windows Unix-like instances, and so on.
Further, as performance needs increase the hardware assets can be improved “under” the OS, so that there is no need for the virtualized OS to know about changes in hardware infrastructure or experience interruptions to reconfigure against changes in underlying technology.
Virtualization offers both a high degree of flexibility and maximizes hardware utilization. Unlike the inefficient silo model, virtualization allows physical assets to be fully used with minimal idle time, lowering costs in acquisition, power and cooling, and maintenance.
But another reason that virtualization is leading the way toward converged infrastructure, said Frey, is that “Most folks looking at production deployment of server virtualization have realized that they need to refresh their data center infrastructure.” In other words, updating the data center for virtualization is already a big step toward a fully-fledged converged infrastructure platform.
A growing range of vendors are now in the business of selling converged infrastructure platforms. At their most unified, these essentially amount to “cloud in a box” solutions, which combine a diverse set of technologies within one set of hardware and managed by a unified set of tools.
HP offers their CloudSystem, a blade-based converged infrastructure platform that combines computing, networking, and storage technologies. It is designed to scale flexibly, so that the virtualized technologies running on the hardware can draw from the whole available hardware pool.
VCE, a combined venture between Cisco, Intel, VMWare, and EMC, markets their Vblock converged infrastructure turnkey platform. Unlike the HP device that combines HP technologies, the Vblock merges technologies from its parent companies.
Dell weighs in with a different take on converged infrastructure, one that acknowledges the existing hardware investments a business may already have. The Dell Virtual Integrated System is a software layer that merges the assets of existing hardware.
And now IBM has entered the fray with their recently announced PureSystem; basically, a cloud enabled data center in a box with integrated management tools and provisioning so customers can expand capacity and get up and running within hours of the boxes hitting the loading dock instead of days or months.
Technology moves quickly and the promise of unified platforms under the converged infrastructure moniker can’t escape this reality. However, adaptable they may be, the only sure truth is that a converged system represents a “snapshot in time” of current technologies. But, as Frey points out, it “remains to be seen” whether “their lifetime will be longer than many of those coming before,” with regard to turnkey unified platforms.
For the business case, though, Frey explains that converged infrastructure solutions “must only prove their worth for the time it takes for them to be fully depreciated, which depending on the organization is typically three to five years. If they can survive and continue to provide value for that length of time, then they have arguably done their job.”
Aaron Weiss a technology writer, screenwriter and Web development consultant who spends his free time stacking wood for the winter in Upstate New York. His Web site is bordella.com.