Amazon Web Services (AWS), the subsidiary that runs the company's Elastic Compute Cloud (EC2), has tinkered with the cloud's established pay-as-you-go pricing structure and come up with what it calls reserved instances. Customers will be able to reserve EC2 capacity and get lower hourly usage charges for what Amazon calls a low one-time non-refundable fee.

The fee, good for either one year or three, applies to each specific instance reserved and works out to a 30% savings over one year and 50% over three.

Users effectively pay for the hardware but not for its maintenance or idle hours and pay usage charges only when actually using the instances. If they don't use an instance, they won't pay usage charges for it.

EC2 general manager Peter De Santis says the company is responding to customer demand. It is closer to the kind of pricing enterprises are used to and assures them of availability. Obviously it helps them establish an ROI, which in turn could push more of them into using Amazon for their base infrastructure rather than just the overflow.

It's also a good gimmick in the midst of a fierce recession when the cloud is already beckoning.

Costs can be split between reserved instances and the conventional demand instances and temporary capacity can always be added. However, a reserved instance bought for, say, a high-CPU extra large instance cannot be changed for another kind of instance.

The reserved pricing is limited to Linux/Unix operating systems in specific data centers, what Amazon calls availability zones, that can't be changed during the lifetime of the contract and for the moment only to the U.S. Amazon expects to extend it to its EU Region in time.

See http://aws.amazon.com/ec2/.