The data center has always been a heterogeneous place, but not for long. Slowly but surely, the variety of microprocessors that used to characterize our compute centers is giving way to homogenous hordes of x86 processors -- mostly from Intel, with a smattering of AMD thrown in for good measure. In and of itself, x86’s increased presence is nothing new -- x86 boxes have long outsold Unix and mainframes on a units-shipped basis. But in 2010, x86 servers represented the majority of server revenue as a whole, estimated by IDC at 66.1% in Q310. Expect that trend to continue, not abate.
The waning days of Unix
The corrolary to x86’s increased data center dominance is the decline of Unix, which data center buyers demoted to a legacy platform in 2010. Oracle, with its purchase of Sun Microsystems in 2009, appears to be hastening Unix’s demise, driving data center managers away with product cancellations and predatory support pricing, although other reports show Oracle/Sun sales holding steady.
Don’t just virtualize, automate!
Of course, part of the reason for x86’s increased data center dominance is virtualization, which has arguably been the defining data center trend of the last decade. But there are signs that all the virtualization hustle and bustle is coming to a close, as IT departments come to the end of the list of workloads they are willing to virtualize. With the bulk of the raw migration work behind them, many IT organizations want to build on the foundational layer they have built, and will introduce automation, a.k.a. private cloud computing, on top of their virtual infrastructure. The hope is to minimize manual tasks such as ensuring compliance, taking inventory, running reports, provisioning new workloads and testing disaster recovery plans. But while it all sounds good on paper, early adopters say they’ve had trouble getting buy-in for automation and private cloud outside the confines of IT.
Storage, data networks converge
Take a look at the average enterprise server, and hanging out the back you’ll find cables going to a handful of Gigabit Ethernet network interface cards (NICs) and a couple of Fibre Channel host bus adapters (HBAs) going to a SAN. That’s all changing with the availability of 10 Gigabit Ethernet (GbE), as data center managers take advantage of the 10 GbE’s superior bandwidth plus new network protocols to transfer both data and storage traffic over a single, tidy link. But while converged networks are the topic du jour for network and storage vendors, plenty of hurdles remain. Most existing storage systems, for one, don’t natively attach to Ethernet, requiring the use of intermediary bridging equipment. Further, most IT organizations aren’t set up for network and storage personnel to work together. An intermediary step toward converged networks may be to take advantage of existing IP storage technologies, like tried-and-true NAS and iSCSI.
Storage is exploding, still
File this under “the more things change, the more they stay the same.” IT professionals continue to grapple with modern applications’ seemingly limitless ability to generate data -- and regulators’ unwavering demand that it be stored ad infinitum. In fact, Gartner estimates that storage capacity in the data center will grow 800% this year. Making matters worse is virtualization, which requires that storage be networked to enable live migration. To get a handle on out-of-control storage growth, data center managers must explore technologies like archiving and deduplication, or risk drowning their data centers -- and budgets -- in a sea of storage.
DCIM tools on the horizon
One of the most pressing concerns for IT managers is extending the life of their current data centers by optimizing power and cooling efficiency. As systems become more complex and variable, XL spreadsheets or Visio diagrams may not be enough to track data center infrastructure components. The entrenched power and cooling equipment vendors launched data center infrastructure management (DCIM) tools that can help data center pros avoid hot spots, or over-provisioning equipment, and other facilities challenges. But third-party vendors have jumped into the fray as well, with sophisticated tools that render data center facilities in 3-D, and include auto-discovery functions that track moves, additons and changes to facility infrastructure, as well as some tools that can even make adjustments to server power draw to improve system efficiency. These are still early days for DCIM, and IT pros are kicking the tires and looking for more standardization, but some companies may not be able to wait and will need more visibility into their facilities infrastructure now.
Colocation data centers shift focus to cloud computing
Last year, colocation data center providers targeted large enterprise data centers. Many companies ran out of data center space and rather than building new facilities, they turned to lease retail-priced raised-floor facilities. In a tight capital market, laying out tens of millions of dollars for a new data center was a hard pill to swallow. This year, colos are chasing the dollars into the cloud. The jury is out on whether cloud computing will save IT managers money, or make cloud providers rich. But the colo companies are poised to cash in. Cloud, Infrastructure as a Service, Platform as a Service and Software as a Service providers need enterprise-class data center infrastructure; they need to be geographically close to their markets to minimize latency; they need access to diverse networks; and these companies need to scale quickly. The smart colos are setting up shop to charge cloud providers a premium.
IT operations deploy more core systems management functions as SaaS
IT cost analysis, help desk, IT service management, even data center infrastructure monitoring -- there are very few systems management software tools that can’t be handled as a service. SalesForce.com and ServiceNow have done a great job providing IT departments the application functionality they need, without the upgrade and maintenance headaches, and even the Big Four are following down that path now, delivering their tools via Software as a Service (SaaS) model. Increasingly, companies are turning to SaaS models because the tools are easier to purchase and deploy, despite the potential for overall higher prices over time.
Data center containers, or pods, take off
Around five years ago, everybody had a good laugh over the Sun Microsystems Blackbox -- a data center Winnebago of sorts. Companies could put them out on their front lawns on cement cinder blocks. Then Microsoft built out a whole data center trailer park and proved the concept could work in specific cases, like giant Web server farm build-outs. But the average data center pro isn’t buying servers in semi-trailer-sized increments, and the idea fizzled. Until now. The second generation of data center containers is smaller and better engineered than their predecessors. SGI’s Patrick Yantz was a part of the Microsoft team that built out the first major trailer installation, and learned from that experience. The SGI’s latest ICE Cube Expandable Modular Data Center is a great example of what’s possible going forward with prefabricated data center facilities.
Cisco UCS grabs mega market share, inspires more converged infrastructure products
Cisco fired the first shot, but now everybody is jumping to the fight for your data center dollars. Some IT managers are calling these converged systems the new mainframes. They comprise servers, storage, networking, operating systems, hypervisor, management tools and middleware all pre-integrated, and each layer of the stack is optimized with the vendor’s secret sauce. Examples include the Cisco UCS, Oracle’s Exadata and Exalogic machines, HP’s Bladesystem Matrix and VCE’s vBlock. These machines can put up impressive performance numbers, and make vendor management simpler, but analysts and IT managers remain wary of vendor lock-in. Cisco has the most to gain in this new market for converged hardware platforms, with installations in nearly every data center around the world.
Check out the rest of our Server Month resources.