Evernote's users have registered over 30 million accounts and uploaded 1.25 billion notes to the service. Each day, Evernote receives hundreds of millions of API requests from over 1000 different applications, and performs advanced processing (like OCR) on millions of notes.
This talk will describe how Evernote evaluated standard "Big Data" software and cloud providers, and why they chose to run their own servers in a simpler "sharded" architecture that has offered linear scalability as Evernote grew. Dave will describe our software architecture and then walk through the way that we've tuned our hardware and operational environment to lower our costs to about $0.05 and 1.4mW per user.