Netflix video quality at scale.

 The Netflix streaming pipeline requires a way to measure video quality at scale. To name a few applications, perceptual quality measurements are used to drive video encoding optimizations, undertake video codec comparisons, conduct A/B testing, and optimise streaming QoE decisions. The VMAF statistic, in particular, is at the heart of increasing Netflix members' streaming video quality. Because of its open-source nature, it has become the de facto standard for perceptual quality measurements within Netflix and throughout the video industry.

The issue of coupling

Video quality measurements were previously generated using our Reloaded production system. This system is in charge of processing incoming media files like video, audio, and subtitles and making them available for streaming. Although the Reloaded system is well-developed and scalable, its monolithic architecture can stifle quick innovation. More crucially, video quality measurements are generated concurrently with video encoding in Reloaded. Because of the tight connection, the following cannot be accomplished without re-encoding:

B) preserving the data quality of our catalogue while we roll out new video quality algorithms (e.g. via bug fixes).

Re-encoding the entire collection in order to generate new quality ratings is a prohibitively expensive method. With our Reloaded architecture, such coupling issues abound, thus the Media Cloud Engineering and Encoding Technologies teams have been collaborating to develop a solution that addresses many of the issues with our previous architecture. This system is known as Cosmos.