Werner Vogels, the CTO of Amazon, delivered a talk to the 12 people who showed up for CS 309 class at Stanford a few days ago… he had previously been a researcher at Cornell for 12 years. Seemed a very down-to-earth guy and talked freely about something I had heard of before but never realized the importance of: Amazon’s services to businesses, including the Simple Storage Service and Elastic Compute Cloud. Why is it a big deal? Because startups (and there was one there that backed his claim up) will never have to deal with infrastructure growth pains again… Amazon spent $2 billion building their data centers, and essentially they’re allowing you to rent virtual machines and storage space on-the-fly, meaning you only pay for what you use, and you can instantly scale up from 1 to a few hundred servers… with prices in the tens of cents/hour, it’s very impressive.
Not to mention Amazon Historical Pricing, which is a statistician’s dream (I spent an entire day trying to find sales & pricing data for pharmaceuticals for a statistics paper, only to find them in a yearly publication known as ‘the Red Book’ in the Med School library).
After the talk, I got a chance to ask him why you could sort by average customer rating with book searches, but not DVDs. He said it was possibly just an oversight, and asked me to email him. Update: This seems to be possible now.
A few months ago I met with my friend Howard Shen who just finished up his MBA @ MIT and asked him about how Amazon’s warehouses work. He explained that there’s no classification system - items are placed randomly on shelves and employees simply scan the item and the item’s location into the database… when an order needs to be filled, you simply look up the number in the database and scan the item to make sure it’s correct. They measure efficiency in terms of how many errors they are between the database and the ‘physical’ reality.