As the premiere search engine in the world, Google must cope with hundreds of millions of searches every day. Understandably, this puts great strain on the few datacenters that Google uses to house their hardware. To remedy this problem, Google has devised a plan that puts more datacenters closer to their clients at a lower cost.

This article by Robert Cringely outlines Google’s two-prong strategy to reducing latency for searches. The first step for Google was to buy up dark fiber. Dark fiber are optical cables that have been laid down by telecom companies for future use but have never been actually utilized for data transfer (they are ‘dark’ because they were never lit up with fiber-optical transmission). In this way, Google is able to create a high-speed communications network for its search engine. The second step is to create the data centers themselves. Google’s ingenious plan involves housing their computer hardware in standard shipping containers. These containers are easily transported (since these containers are primarily used for shipping) and compact. Google envisions detecting areas of high network load, shipping these containers to a parking garage in the area, and then hooking the container up to electricity, cold water (to dissipate heat from the hardware) and the optical data link. Load-balancing software could then redirect many queries to this new on-site datacenter, greatly reducing overall network load.

In general terms, this strategy would allow Google to selectively place nodes where they are needed the most. This also saves money over building permanent structures which could not be moved once the load subsides. Google is a world leader in providing web applications, such as GMail, Google Maps, and Google Docs, so load balancing is a forefront issue. As more and more software moves from the desktop into cyberspace, such distributive strategies will become increasingly necessary.