Massively Parallel Open Source Encoding for Adaptive Streaming

Encoding of premium UltraHD content in adaptive streaming ecosystem trades the number of encode jobs for a wider network and device reach. The recent emergence of content and context adaptive technologies which often require “trial encodes” further raises the amount of computational resources needed to encode a single video. — Distributed encoding, where non-overlapping chunks of the same video are encoded in parallel on different machines, emerged as the mainstream response to these challenges. This approach reduces the overall encoding time and makes the best use of available hardware resources, and is an excellent fit to a cloud or cluster environment. — In this paper we will describe an implementation of the massively parallel approach to video encoding. We will start from defining a chunk encode, “joblet”, and what it entails. Lastly, the paper will demonstrate a dramatic increase in the computational efficiency of high-quality offline video encoding without a significant impact on video quality.

Published

2018-10

Content type

Original Research

Keywords

Video compression, distributed encoding, distributed processing, HEVC, x265

DOI

10.5594/M001829

ISBN

978-1-61482-960-7