We will plant a tree for each of the first 5,000 downloads.
Call it a private cloud or convergence or a way to streamline staffing levels or a recipe for vendors to lock you into their proprietary systems--whatever your take, it's the future of data centers large and small.
Most of the midsize companies we work with have embraced server virtualization with good results. But they're hitting a bump on the way to private cloud nirvana: Storage is still application-specific. There's a SAN for the Exchange server cluster, a different SAN for the Oracle database, another for the Web cluster, and a big NetApp filer for unstructured data. Servers may be virtualized, but moving one from the Exchange cluster to the Oracle RAC cluster isn't easy.
The promise of a private cloud, of course, is that once we do away with application-specific storage (and servers and networks), everything melds into one resource pool that you carve up to deliver IT services. But even with advances in virtualization--server, storage, I/O, and network--merging systems is complex, expensive, and disruptive, especially for cash-strapped midsize companies.
We need a staged approach.
Beyond The Flash
First, forget the buzzwords. Stop worrying about converged this and unified that, and get back to the basics of why you must do this--it's because a well-designed private cloud provides substantial benefits. And whether network jockeys and Fibre Channel gurus like it or not, islands of storage are not the future. Standards bodies are gaining consensus on Fibre Channel over Ethernet. Switching vendors are becoming more virtual-machine- and virtual-I/O-aware. Intel and AMD are packing more cores onto each physical CPU socket. I/O vendors, such as Brocade, Emulex, and Intel, are building the next generation of flexible I/O chipsets, and server vendors are cramming more I/O onto system boards.
There's a lot at stake for vendors, because, let's face it, commodity hardware profit margins are slim. The Ciscos, Dells, and Hewlett-Packards of the world aren't lighting up Wall Street with revenue from the standard five-year hardware refresh cycle. These and other vendors are setting their sights on selling unified stacks to midsize companies with data centers badly in need of updating.
As for software--where the real money is--the big opportunity is in next-generation management software that will effectively control our highly virtualized data centers.
Lay The Groundwork
Three broad capital costs underpin any private cloud project. The first is the server virtualization platform. "Network convergence" is really just a fancy term for "I/O virtualization," and we can't virtualize I/O without an underlying server hypervisor. Our recent InformationWeek Virtualization Management Survey shows 39% of nearly 400 respondents from companies with 50 or more employees expect to have upward of 75% of their production servers virtualized by the end of next year. That's pretty good, but as we move toward pervasive server virtualization, the need for I/O consolidation will grow, because you can support only a finite number of VMs over a single gigabit link.
The result of a growing virtualized environment is a mishmash of disparate interfaces that lack flexibility and introduce lots of independent points of failure. The complexity and cost of managing multiple switch fabrics is public enemy No. 1 for many SMBs, but that problem must be weighed against the cost required to consolidate I/O properly.
Google in the Enterprise SurveyThere's no doubt Google has made headway into businesses: Just 28 percent discourage or ban use of its productivity products, and 69 percent cite Google Apps' good or excellent mobility. But progress could still stall: 59 percent of nonusers distrust the security of Google's cloud. Its data privacy is an open question, and 37 percent worry about integration.