TikTok video pre-loads cause 'massive data wastage'

"News of that finding was the subject of a paper [PDF] titled "Wisely Optimizing Short Video Streaming for a User-Vendor Win-Win Outcome" presented at the SIGCOMM academic networking conference a couple of weeks ago. The paper's authors work at Columbia University, Renmin University of China, Tsinghua University, and ByteDance. Their work opens by observing "Short video streaming platforms widely employ video prefetching to ensure users' quality of experience (QoE), but frequent user swipes lead to massive data wastage, creating a significant financial burden for vendors.""

"The authors think those discards account for "over 40 percent of the streaming vendors' data transfer costs" and suggest they need to change the way they prefetch videos to save money. The paper suggests existing alternative QoE mechanisms can't help, because their developers designed them for conventional video on demand services. The authors therefore developed their own QoE framework."

Short-video platforms commonly prefetch video chunks to improve perceived quality of experience (QoE), but frequent user swipes cause large volumes of prefetched content to be discarded. Study of Douyin and Kuaishou found both services prefetch the first chunk of the next few videos, which can lead to video stalls and significant wasted transfers. The discarded prefetched chunks may account for over 40 percent of streaming vendors' data transfer costs. Existing QoE mechanisms for video-on-demand do not address this behavior, so the researchers developed a new QoE framework aimed at guiding prefetching to reduce waste and costs.

Read at Theregister