Ultra-High Resolution Video Distributed Transcoding System Using Memory-based High-speed Data Distribution Method

Delivering ultra-high resolution video in realtime over a public network is a challenging task. There are two possible solutions: 1) tiled encoding and streaming technology can effectively transmit the video with an extremely large bit rate; and 2) distributed transcoding can encode ultra-high resolution videos in real-time. Distributed transcoder-based real-time encoding of ultra-high resolution videos has latency due to input stream distribution and output stream integration. The existing disk-based distributed transcoding system has a limitation on the data transfer speed between the master and the slaves due to the read/write speed of the disk. In this paper, we propose a tiled distributed transcoding system using memory-based distribution method. The proposed system supports High-efficiency Video Coding (HEVC) or AOMedia Video 1 (AV1) transcoding on Spark on Yarn platform. Experimental results show a transcoding speed improvement of about 21% compared to the existing system.