‘Blades on Wheels’ — Plug ‘n Play Data Centers

Some of the major IT vendors have taken a new tact in competing for the heart and soul of the data center. They’re offering densely packed data centers-in-­a-box that literally can be rolled up to any corporate location. Could this be the answer to the burgeoning space and power requirements that are said to be afflicting many growing data centers? Or is it an example of vendors running in the wrong direction?

There’s nothing new about the idea of mobile data centers. For the past decade or so, disaster recovery vendors have offered fully contained, trailer­based mobile data centers that could be rolled up to customer sites to provide con­tinuity to disrupted IT operations. Now, companies such as Sun, IBM, Dell, Rackable Systems, and even Microsoft are proposing that containerized data centers be pluggable into enterprises for any purpose – such as expansion – on short notice. Think of it as a very large “blade” on wheels.

The growing prevalence of commodity hardware and software make this eco­nomically feasible, said Microsoft’s James Hamilton in a recent position paper (Word document). “These commodity clusters are far less expensive than the systems they replace, but they can bring new administrative costs in addition to heat and power-density chal­lenges…. [Hamilton proposed] a data center architecture based upon macro-mod­ules of standard shipping containers that optimizes how server systems are acquired, administered, and later recycled.”

As Hamilton put it, a standard 20x8x8-foot ship­ping container is “ideal” for this purpose, since not only is it is rugged and built to withstand ocean voyages, but also “relatively inexpensive and environmentally robust.” Upon delivery to a site, a data center container could simply be attached to the network, chilled water, and powered up. Each container can be fully equipped with networking gear, compute nodes, and persistent storage. Hamilton even went as far as to predict that these plug-and-play data centers won’t even need any main­tenance or servicing, since they will be loaded with redundant components – “the entire module just slowly degrades over time as more and more systems suffer non­recoverable hardware errors. Even with 50 unrecoverable hardware failures, a 1,000 system module is still operating with 95 percent of its original design capacity.”

Sun Microsystems – a company always trying its best to shake up the established order – already unveiled its own container-based data center. Last October, the ven­dor announced “Project Blackbox,” which is scheduled to be made available by mid­year.  The current prototype could support about 250 servers, up to 1.5 petabytes of disk storage, 7TB of memory, and support up to 10,000 simultaneous desktop users. The interior resembles the inside of a space station. One observer remarked he was half expecting the system to utter “Hello, Dave” when he walked in to take a tour – à la the interactive HAL 9000 supercomputer of 2001: A Space Odyssey.

The economies of scale – mass-produced data centers in a box – are compelling, but in the age of emerging software-as-a-service, on-demand, and in-the-cloud com­puting approaches, is it really needed, since companies may increasingly turn to out­side resources for IT requirements anyway? Proponents of containerized data cen­ters say that trends such as SaaS and Web 2.0 are actually what are driving the mar­ket for these solutions. Companies that provide such services constantly need to keep beefing up their data centers to take on new customers. However, beyond the service providers themselves, how much of a market will there be?

A trend going in the opposite direction is the increasing compute power constantly being packed into smaller and denser chips, as well as geometrically growing stor­age capacity that shows no signs of abating. It’s not inconceivable that all the func­tions of an average data center could eventually be packed into a box the size of a desktop computer. Consider all the scientific and high-end applications that once required Unix computers that now can be run on a laptop.There’s also IBM’s con­solidation pitch, in which thousands of distributed servers could be virtually replaced by a single System z mainframe.

If a company can get a refrigerator-sized mainframe box to support most of its business, why would there be a need to ship in an entire trailer of additional computing?

You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.

2 Comments »