It was a dark and cloudless night. No, this is not the start of a cheap paperback novel. It's about the prehistory of cloud computing, long before Amazon's S3 storage cloud went down last month.

Over a decade ago I worked at Open Text, which was building one of the first Web search engines. The Web index ravenously ate up storage, and we were buying 9 gigabyte hard drives (the biggest available at the time) as fast as we could. And we were learning a hard lesson in mean time between failures (MTBF).

Fast forward to 2008. Now I can have all of the data storage I need as I need it for a fraction of the cost. Only I don't have to purchase it, or maintain it. I also don't really have any control over it. And it occasionally does go down. But so would a data center I managed myself.

Once upon a time, a data-intensive startup such as Google would have needed to buy hardware to build a server farm or two. They would have to ensure reliability, and scale quickly to meet anticipated user demand. Server farm expansions were risky and sometimes fatal for '90s-era startups, as these became wasted investments if demand turned out to be feeble.

A startup today just points at a service such as Amazon S3 and starts to store data. As user demand increases, the space is there to accommodate it, and they are charged $0.15 per GB per month (plus in/out charges). If they've made a horrible mistake and there are no users, and therefore no data to store and transfer, they don't get charged.

Wonderful, you say, but that's just data. Startups still need servers. Enter Amazon EC2 -- cloud computing -- providing virtual servers that you can configure and use as you need them.

It isn't just Amazon either, though they seem to have been the first to provide such services. IBM, EMC, and others are jumping into the ring -- sorry, cloud -- too.

While there may be a few hiccups along the way, such as Amazon's massive service outage last month, the value of a scalable system costing just pennies per gigabyte without additional maintenance overhead is hard to argue with.

The potential impact on Internet startups is tremendous. Line items that used to eat up millions of dollars drop to almost nothing. Entrepreneurs and their staffs can let someone else deal with infrastructure worries, and concentrate on what really matters – developing products, interacting with customers, and taking advantage of market opportunities.

Related news, commentary, and predictions:

• Prediction: Microsoft to announce online versions of Office
• Larry Borsato: Communications: Why do we accept less than 99.999%?
• Larry Borsato: ISPs, the music industry, and the Web
• Larry Borsato: If you want a job, we don't want you
• News: Outage hits Amazon S3 storage service
• News: Microsoft dings Amazon's S3

Note: Anonymous comments on The Industry Standard are disabled. To leave a comment and participate in the Standard's prediction market, please register first.