Dissecting the performance of production QUIC
Content Delivery Network (CDN) administrators are often faced with a fundamental question: “Should I adopt a new shiny protocol or should I continue using a trusted and reliable protocol?” This has especially been true with the development, release, and standardization of QUIC.
Unfortunately, academic publications and industry experiments provide conflicting results, because of inappropriate baselines and unstandardized testing environments, making it challenging for CDN administrators to make informed decisions.
We at Brown University observed that bridging the chasm and providing consistent results is impossible using current methods. In particular, academic benchmarking on open platforms lack key optimizations, and industry benchmarks overlook proprietary optimizations.
In a recent paper published at the Web Conference 2021, we addressed these issues by providing a testing harness and a method for dissecting the performance of production QUIC. This post shares some highlights from our study.
