What People Get Wrong About Leverage, Scale, And Growth

In the early days of Amazon, the company handled increased page requests by purchasing and installing larger, more powerful servers from Sun Microsystems. These machines were incredibly expensive but represented the best technology available at the time. A web page request requires a certain amount of CPU, memory, and other resources. The approach to scaling at that time was to buy hardware with the fastest CPUs and the most memory. Today, we call this approach vertical scaling.

During Amazon's earliest period, the website was served by one of eight Sun machines located in downtown Seattle. My friend Dave Glick shared an interesting anecdote from a late Thanksgiving night in the late '90s. They needed to move the website from downtown to their first data center in the industrial district, but they weren't sure which machine was serving the website. To identify it, one person continuously reloaded the website while they powered down the servers one by one. When the website finally would…