Strange Loop

The Evolution of the Flickr Architecture

Flickr was created over 6 years ago - an eternity in tech years. While the site's general architecture has has held up fairly well, some pieces have started creaking at the seams. A new crop of engineers has started evolving the site's architecture to take advantage of new technologies. We'll discuss what the Flickr team is doing to keep up with the times while operating a quickly-aging, labyrinthian codebase. Topics will include, but will not be limited to:

• MySQL
  • why it's still our hammer - some numbers
  • the tools we use to make managing MySQL easier
  • knowing when it's not the right tool for the job
• Redis: picking up MySQL's slack
• PHP
  • testing in a legacy, procedural codebase
  • introducing objects and exceptions without breaking things
  • Java: the dark horse; how we use various specialized daemons to help scale the site
  • Hadoop: crunching the numbers
  • Frontend performance: Going from 5 seconds to below ~1
  • Deploying: we still deploy ~10 times a day and it's still awesome/li>


 The topics will be accompanied by appropriate war stories. Come hear what a new generation of engineers is doing to keep Flickr running smoothly as it continues to grow. If you're not interested in that, come laugh at the tales of all the unbelievably dumb things we've done over the years.

Mikhail Panchenko