Low-Cost Video Transcoding at the Wireless Edge

Abstract

With the proliferation of hand-held devices in recent years, wireless video streaming has become an extremely popular application. Internet video streaming to mobile devices, however, faces several challenges, such as unstable connections, long latency, and high jitter. Bitrate adaptive streaming and video transcoding are widely used to overcome such problems. However, there are still several shortcomings of these approaches. Bitrate adaptive streaming cannot provide fine-grained adaptation. Moreover, video transcoding is expensive and is hard to apply for live streaming. In this work, we propose to use a low-cost video transcoding solution running at the wireless edge, such as home WiFi Access Points (APs). Instead of having expensive server-based transcoding, we designed and implemented a real-time video transcoding solution on a low-cost hardware, Raspberry Pi. By running our transcoding solution at the wireless edge, it can provide more agile adaptation to sudden network dynamics and is able to incorporate clients' feedback quickly. Our transcoding solution is transparent, low-cost, and scalable. Thus it allows broad and quick deployment in home WiFi APs (and also in cellular base stations). The evaluations reveal that our system enhances the performance of video streaming compared with other bitrate adaptive solutions. It provides higher video bitrates (at least 2.1×) without causing video stall or rebuffering.