Facebook Developers Push Out New Tools at @Scale Event

At Facebook’s annual @Scale Conference 2017, the social network announced several developments that solve some complex technical challenges, including around artificial intelligence and image creation.


Facebook keeps adding new users (2.01 billion by last count), building new data centers (it’s now up to 10) and its desktop and mobile advertising businesses are going great guns. The market cap is nearing $500 billion and Mark Zuckerberg is preparing again for paternity leave, so one could say all is going pretty well there at its headquarters in Menlo Park, Calif.

Outside of knowing that the bills and salaries are always going to be paid, Facebook’s software and hardware development teams don’t need to think about all that business stuff. They simply go about their business each day, starting and maintaining numerous initiatives as usual behind the frontline businesses. They've also been a valuable resource in the open source community for the last decade or so.

At Facebook’s annual @Scale Conference 2017 Aug. 30 in San Jose, the social network announced several developments that solve complex technical challenges and ultimately scale the technology driving current and future company innovation for the billions of people that use Facebook every day.

Announcements and important data points include:

  • Continuous Push: Earlier this year, Facebook moved 100 percent of its production web tiers to a continuous deployment model, so Facebook ships tens to hundreds of code changes every few hours, 24/7, 365 days per year. Benefits here include the ability to ship fixes more quickly; it empowers a global development team to ship code whenever best for them, resulting in stronger tools/processes to handle the frequency of releases and better preparing the company to handle future growth; a real-time release and feedback system.
  • Overall, this model improves both product development and product stability. Continuous deployment enables fast iteration on new features, as well as improved safety and stability for people using facebook.com.
  • Continuous deployment works at Facebook scale. The team found that shipping more frequently doesn’t affect productivity or quality; the amount of code shipped scales with the number of developers and the rate of issues arising remains constant.
  • Advanced 360 Photo Optimization: Facebook shared how it re-engineered its photo infrastructure to allow for immediate hi-res viewing of 360 photos without causing other content in News Feed to stall. This comes on the heels of the company’s new in-app capture experience and its approach of allowing for “tiled” storage and serving of photo content is similar to how mapping technologies deliver aerial imagery.

Facebook also released new research on how deep neural networks can be used to correct orientation of a crooked 360 photo. Editing this kind of rotation is straightforward for traditional 2D photos, but the same types of tools are not widely available or intuitive for immersive 360 photos. The deep neural network the team developed automatically fixes rotated scenes.

  • LogDevice:  Facebook’s custom-built log storage system, LogDevice, helps ensure that Facebook's 10 global data centers can efficiently replicate log data between distributed data stores, while maintaining high-write availability, durability, and consistency—among other things.

Check out Facebook’s Engineering Blog for updates on these and other projects.

Chris Preimesberger

Chris J. Preimesberger

Chris J. Preimesberger is Editor-in-Chief of eWEEK and responsible for all the publication's coverage. In his 15 years and more than 4,000 articles at eWEEK, he has distinguished himself in reporting...