Using The Power of Edge Compute To Enhance Streaming Delivery
summary
Edge code converts existing HLS on the fly—convert MPEG-TS HLS to fMP4—and adds Roku JPEG trick‑play without remuxing the whole library, working for both VOD and live. Apps run on Compute@Edge using Rust and FFmpeg compiled to WASM/WASI, push output into CDN caches, and use pre‑fetch/per‑warming to avoid origin fetches. Video decode dominates cost (about 66% of processing time), but the platform can do thousands of JPEGs per second and scales instantly while keeping stateless, just‑in‑time workflows.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:28Why Build Dynamic Edge Applications?
- 1:21Dynamic Edge Application Characteristics
- 1:54Application Environment
- 2:24Why Edge Computing?
- 2:54Segment Pre-Fetch
- 3:27Request Handling
- 3:49Rendition Playlist
- 4:20Pre-Fetch Operation
- 6:24JPEG Trickplay
- 7:20Master Manifest
- 9:14Using FFmpeg CLI To Generate JPEG
- 9:39Things To Consider
- 10:36Observations
- 11:12Stream Format Conversion
- 12:24Playlist Manifest
- 14:41Conclusion
related talks
16mImproving Video Delivery and Streaming Performance with...
41m8K Codec Low-Latency Livestreaming, and AI Streaming
30mOther browser APIs that are going to change video on the web
10mEdge Compute
43mModern Streaming Video for the Jamstack
20mDecentralizing Video Streaming
12mExploring Video Editing in the Browser
10mEnhancing CDN Performance and Cutting Egress Costs in Large Video Libraries