Web STD

Posted June 25, 2008 at 10:16 am by Eelco van Beek in Infrastructure.

Some weeks ago at the office we were honoured by the visit of Werner Vogels, CTO at Amazon and an STD expert par excellence. It was very interesting to learn about the motivations of a business to make STD its core business. Each page within Amazon.com is the result of tens of services, each one taking care of one part of the page. Why was this method chosen?

From an infrastructural perspective the benefit is that each component can be individually scaled, based on demand of that specific component. So, cost wise much more interesting than scaling a complete, monolithic application. As the whole architecture is service-oriented, certain functionalities could be purchased somewhere else (for example ad function from Google). Another benefit this approach brings is less complicated service security and maintenance. In case of detected bugs, these are resolved in one central location, and the same goes for security errors.

At some point, Amazon decided to commercially exploit STD, and not without success. Many of Amazon’s own used services are against a fee per usage. Possibilities are endless and include storage, virtual computers, logistic processes and payment functionality. Infrastructure components must be designed on the basis of expected peaks. By hiring these components as services, or by using, for example, virtual systems on which such components can run, it is possible to practically build without investment.

The only remaining question is who dares to adopt external services as business critical components in architecture.