IBM to build 'cloud computing' data center |
 |
By Mark Fontecchio, News Writer
04 Aug 2008 | SearchCloudComputing.com |
|
|
IBM has announced that it will spend $360 million to build a new data center at its campus in Research Triangle Park, N.C., that will be designed to handle cloud computing.
In the annoucement, IBM stressed certain features of the new facility, including energy efficiency and high-density computing. But in fact, these are common components of all of today's enterprise data centers and are hardly differentiators of a cloud computing-oriented facility. So what distinguishes a true cloud computing data center from a regular enterprise one?
According to Charles King, an analyst at Pund-IT Inc., there are three main features of a cloud computing-oriented facility: commoditization, scalability and fault tolerance for large numbers of servers.
"In traditional data centers, companies tend to deploy individual systems that have a lot of redundant availability features: redundant power supplies, redundant power and cooling," King said. "While those features can be really valuable on an individual server basis, when they're replicated over tens of thousands of servers, they become incredible sources of inefficient power usage."
"With cloud systems, the basic idea is that you throw a huge number of servers at an application and automatic provisioning software, from a pure statistical standpoint, if a server goes down, you simply roll that workload over to the gazillions of other servers that are in the data center," King added.
IBM's 'cloud computing' facility
Mountain View, Calif.-based Google Inc. has built its data centers consistent with King's definition. Google's data centers feature simplified x86 servers, the ability to add hundreds more, and when servers break down, the use of such scalability to roll workloads over to available hardware.
So the question is, does IBM's new data center facility qualify as cloud computing? The 130,000-square-foot facility will include 60,000 square feet of raised floor, and according to IBM Director of Advanced Customer Solutions Jay Subrahmonia, it will include System x, p, and z servers, which encompass its x86, Unix and mainframe platforms. The roster of hardware conflicts with King's definition of a cloud-computing data center. According to King, "while you would likely see multiple generations of Intel or AMD processors in a single cloud facility, unless you had specific applications that could only run on Unix or the mainframe, most would be x86 systems."
But Subrahmonia countered that IBM views cloud computing as comprising a wide range of IT services that an end user can access without having to worry where the services come from.
"It includes a user interface that masks the complexity of the underlying infrastructure," she said. "Really it's about an IT infrastructure that can deliver services to any Web-connected user."
Although IBM's North Carolina data center might not fit King's definition of a cloud-computing data center when it comes to x86 server commoditization, IBM's approach to scalability does converge. In particular, Subrahmonia said that when demand warrants it, IBM will use virtualization on its System x servers, including Citrix Xen technology, to scale up infrastructure even if it is virtual infrastructure.
Subrahmonia added that large amounts of data and transfer speed are other key attributes of IBM's cloud-computing data center as well as lots of network bandwidth and massive storage capability. IBM, however, was unable to provide details about the data center's network and storage infrastructure, because some aspects of facility design are still in the works.
Let us know what you think about the story; email Mark Fontecchio, News Writer.
|
|
|
|
