By Christine Martino

Despite all the excitement about clouds, only a handful has yet been built, mostly because infrastructure is such a challenge.

Take Google. When the company pioneered what might be called the first cloud, it couldn’t find the streamlined, low-cost servers it needed, much less in the numbers it demanded. So, Google built its own.

Before cloud-type systems can bloom, businesses need ready access to the kind of infrastructure Google had to cobble together itself: high scale, lowcost and mass customized. These extreme scale-out infrastructures will not only seed clouds, but nourish Web 2.0 and high-performance systems while inspiring business models we can’t yet imagine.

Scale is the heart of clouds and similar systems. They start big and only grow bigger. One example: Facebook’s 175 million users¹ upload 850 million photos and 8 million videos every day.²

Yet, few traditional IT architectures can scale effectively to thousands of nodes. Fewer data centers can support that number. And fewer vendors have the IT supply chain or deployment options to effectively deliver it.

An extreme-scale infrastructure has to be built to deploy, scale and maintain thousands of nodes at a time. That means servers that can be added or replaced with plug-and-play ease. Cabling that’s easily accessible and racks capable of packing in more components in less space. All these elements must also run (and be cooled) at the absolute lowest possible power consumption.

Where other businesses count dollars and days, extreme-scale ventures count pennies and minutes. With razor thin margins there’s no room for an ounce of waste in the infrastructure. Servers must be streamlined to their essential components and features. That means not only cutting out the extras, but also rethinking the basics, including removing metal and materials that add shipping weight and take up data center space.

Energy efficiency, always important in data centers, becomes critical in the extreme scale-out environment. Low power use and energy-efficient design must be integral to the fabric of the infrastructure. Technologies such as shared power supplies, temperature sensors, variable power fans and intelligent environmental controls are keys to keeping power use, and thus costs, low.

Services, too, have to make sense. While an enterprise may pay top-dollar to assure server uptime, the fault tolerance for cloud-like systems is often built into the software. If a server fails, getting it repaired and back online isn’t an emergency. Periodic server swaps or having a stocked parts closet on hand are more appropriate, and less expensive, in this environment.

Rolling out new services and projects fast, cheap and at a competitive scale requires a vendor who has supply line and delivery mechanisms built for the job.

Mass customization and just-in-time delivery aren’t just buzzwords in the world of extreme scale. The ability to deliver large-scale orders to exacting configurations anywhere in the world, faster and with less overhead can mean the difference between success and failure in an extreme scale venture.

As extreme scale infrastructures take shape, clouds and other emerging business models that depend on massive computing architectures will bloom and grow. This radically-different IT infrastructure will bring ideas and applications once thought impossible into reach.

And that’s surely just the beginning. Just as the Internet sparked ideas unimaginable before its existence, extreme scale-out environments will doubtless pave the way for new business models in the cloud, in Web 2.0, in high-performance computing or anywhere an audacious idea takes root.

	Your feedback is important to us. Was this article useful/informative?
	Not at all(1)	—	Neutral(3)	—	Definitely(5)