Uploading Instagram Videos with 94% LESS CPU

Now it's easier to spend even more time on the platform

tldr;

Instagram used to compress the same video twice for all users (progressive and adaptive bit rate).

They realized they could save time by…

1. first compressing the video to progressive encodings

2. Then convert that progressive encoded video into an adaptive-bit-rate video.

Results: The entire process uses 94% less resources.

---

🙋‍♀️ Credit to The Original Author

Ryan Peterman

Wrote the awesome original article this was summarized from on improving Instagram compute time. He’s an amazingly talented engineer and you can learn more about his experience and senior+ skills at

The Developing Dev

🤔What’s the problem?

Instagram needs to maintain video formats for all users.

The formats include.

• Minimum Functionality type Progressive encodings for all older Instagram apps.

• Minimum Functionality type Adaptive Bit Rate for older Instagram apps that can support bit rate streaming.

• Advanced Functionality type encodings for high quality videos for newer Instagram apps.

Minimum functionality type encodings take most of the CPU resources where in 1 year no infrastructure could scale to support this growth.

⚙️Requirements

• Continue to support progressive encodings for older Instagram devices. This means to be backwards compatible.

• Continue to support adaptive bit rate encodings for older Instagram devices. This also means to be backwards compatible.

• Scale and extend advanced encodings for higher quality videos for devices that can play the videos. Give users the best performance if they can handle it.

• Minimize resource consumption for progressive or adaptive bit rate encodings to scale. Be efficient with your CPU usage.

🚧 The Big Design