48: A 94% reduction for basic video compute time on Instagram

Meta Tech Podcast

Ryan and his team found a quick way of reducing the compute resources spent on encoding videos for Instagram by 94%, but that was actually the easy part. Tune in to learn what the fix was and how you roll out changes that can affect the user experience of billions of users.

Got feedback? Send it to us on Twitter (https://twitter.com/metatechpod), Instagram (https://instagram.com/metatechpod) and don’t forget to follow our host @passy (https://twitter.com/passy and https://mastodon.social/@passy). Fancy working with us? Check out https://www.metacareers.com/.

Links:

• Reducing Instagram’s basic video compute time by 94 percent - Meta Engineering Blog: https://engineering.fb.com/2022/11/04/video-engineering/instagram-video-processing-encoding-reduction/
• The Diff: https://thediffpodcast.com/
• Unix Signals in Production - Dangers and Pitfalls: https://developers.facebook.com/blog/post/2022/09/27/signals-in-prod-dangers-and-pitfalls/
• Introducing Velox: An open source unified execution engine - https://engineering.fb.com/2022/08/31/open-source/velox/

Timestamps:

• Intro 0:06
• Intro Ryan 1:40
• Transcoding Video at Instagram 2:52
• Codecs and Tradeoffs 5:33
• Client Support 7:13
• Where did the compute go? 9:15
• ABR 10:59
• Progressive/Non-ABR Encodings 12:31
• Saving Encoding Time 13:10
• Testing the Changes 17:39
• Results 26:43
• Popularity Predictions 28:32
• Outro 36:31