Infrastructure for rent, going cheap

I heard a lot of interesting things at Roman Staněk’s presentation on Monday. What stayed with me most was the idea of renting infrastructure. It makes sense. Infrastructure is a service, so why not rent it? Lately Amazon has been getting the most attention as a provider of virtual computing power, data storage and a database.

Amazon has launched three interesting services, Amazon Simple Storage Service (Amazon S3), Amazon Elastic Compute Cloud (Beta) and Amazon SimpleDB (Beta). Let’s imagine how rented infrastructure would work for a software company like yours.

Our company F has source files it needs to build. It rents Amazon’s S3 service for physical storage. It also rents Compute Cloud and supplies an image with a source control system such as SVN and a CI server already installed. It uses SimpleDB to store the build results. What does the company get out of this?

• It doesn’t have to own the hardware for build machines or have a department to look after it. There’s no point in having build machines sit idle over the weekend, for instance. When demand spikes, say as a release approaches, the company can rent more capacity.
• The provider takes care of service reliability, including data backups and machine failover, as well as security. These aspects of the service are usually covered by a service-level agreement, so they’re contractually guaranteed.

Building that sort of infrastructure on your own takes more than a week or a month, assuming you’re doing more than cobbling software together. Renting services also changes the staffing equation. You no longer need people or a whole department to look after the infrastructure for you.

A software company should focus on developing software. Spending its resources building infrastructure will hardly yield the same quality and efficiency as hiring someone who does nothing else. These services are still a long way from wide adoption, of course. I doubt it’s a technical problem. It’s a question of making that proverbial mental leap, or waiting for someone else to make it and following if it works.

I have no doubt Roman Staněk has made that leap…