Tim O’Reilly's conversation with Jeff Bezos and conference welcome at Web 2.0 Expo explored the ideas behind Web 2.0 and Amazon's Web infrastrucutre services.
Tim O’Reilly
- Web 2.0 is happening and we are just at the beginning. It’s equivalent to the introduction of VisiCalc during the PC era. There are a few killer apps and we are starting to see a lot of competition emerge.
- Consolidation & failures will happen but the ideas behind Web 2.0 are bigger then any single company. It’s about figuring out how to leverage the global collective knowledge of all people.
- The future is not like the past. Early practitioners are often shaped by the imagination of what came before. Example: horseless carriage for cars.
- The Web is special because it can harness collective intelligence inline any other medium. It also provides new business advantages in the form of data collection.
Jeff Bezos
- When you start a new business, you want to go as fast as possible from idea to successful product.
- Between the idea and what you want is a lot of heavy lifting: server hosting managing bandwidth, moving facilities, contract negotiation, etc.
- Often you have to loop through that cycle. Those who cycle the loop the fastest are the winners. It’s a high price of admission.
- For this need, Amazon introduced their infrastructure Web services: SQS, Mechanical Turk, S3, and EC2. They are built to manage the heavy lifting.
- S3: Storage in the cloud. Accessed through Web service APIs, variable pricing, eliminate fixed costs and fixed burdens. 5 billion objects within the system today. Peak day: 920,847,345 requests; Peak second: 16,607 requests
- Services orientated architecture: Web-scale architecture needs loosely coupled pieces. If parts get slow, the whole application does not grind to a halt.
- The key to a loosely coupled architecture is a queue service.
- Consumer (the services that carry out operations) queue can clone itself. It builds new hardware to increase throughput when the producer (teeing up work) queue gets long. Elastic computing: can increase or decrease based on needs.
- Amazon: can we package things that we are good at the right way? Leverage the infrastructure powering Amazon and make it a service.
- Capacity constrained is not good way to run a business. Its better to be demand constrained.
- Base strategy on what will not change over the next 5-10 years.
- Physical movement of goods will not change in next ten years. Amazon has fulfillment infrastructure service: pay by the cubic foot. Can control storage & shipping via Web service APIs (virtual warehouse).