Actually the keynote (and the interview) was very interesting, some points Dr. Vogels discussed where:
- The problem of state management ("state management is a dominant factor in scaling")
- Amazon as company: "Amazon is a technology company that accidentally works as a retailer"; he also shows a series of other E-Commerce sites like Marks and Spencer, Mothercare, Smug Mug and others that are actually build on top of Amazon technology.
- Amazon apparently goes the way (similar like EBay): from a retailer (auction house) to an e-commerce technology provider/platform.
- A dominant issue in the talk was scalability. COTS products typically do not scale the way Amazon needs it (we tried out mainframes - for one year). Vogels refers to the stability and self-organisation features of biological systems and names particularly "Apoptosis": cell-death; allthough daily 50-70 billion cells die every day, the biological system is stable, aka the human stays alive.
- Thus Amazon services are build highly redundent. The loss of a complete datacenter would not harm the customer experience. He additionally shares two experiences that might contradict certain academic ideas:
- "Everything fails, all the time"
- Systems do not fail by stopping, they might actually do all sorts of weird things in between.
- Vogels claims, that Amazon did SOA before it became a buzzword.
- So eventually his bottom line is "Architecture for change".
Check out the podcast website!
Or go directly to the feed page.
So, to eventually answer the question in the title, a last quote I personally liked: "Does Amazon do research? We call it production."