Fixing the performance of your (probably broken) HTTP/2 deployment
Patrick Meenan lets you in on how HTTP/2 prioritization is effectively broken in most deployments and shows you how to detect, debug, and fix the issues.
| Talk Title | Fixing the performance of your (probably broken) HTTP/2 deployment |
| Speakers | Patrick Meenan (Facebook) |
| Conference | O’Reilly Velocity Conference |
| Conf Tag | Building and maintaining complex distributed systems |
| Location | San Jose, California |
| Date | June 11-13, 2019 |
| URL | Talk Page |
| Slides | Talk Slides |
| Video | |
Working HTTP/2 prioritization is critical for web performance. Even though most servers, content delivery networks (CDNs), and load balancers technically support HTTP/2 prioritization, it’s effectively broken and has no practical effect. When prioritization is broken, the impact on end-user performance can be dramatic and lead to tens of seconds of delays in rendering the critical content on the page. Patrick Meenan shares some background on how he initially stumbled on the problem with Cloudflare’s HTTP/2 deployment, the root causes of the issue, and how Cloudflare fixed it. That was just the tip of the iceberg, though, and led to building a test suite that can test for working prioritization. Every deployment he tested was broken. You’ll use the test suite, find known issues, and learn resolutions for various platforms and how you can detect and resolve the issue with your own deployments. Finally, Patrick touches on the state of prioritization support in the browsers and upcoming changes with HTTP/3 and how that will impact the issue (or not).
