Abstract
The rapid growth of video-sharing platforms has driven immense storage demands, with disaggregated cloud storage emerging as a scalable and reliable solution. However, the proportional cost of cloud storage relative to capacity and duration limits the cost-efficiency for managing large-scale video data. This is particularly critical for cold videos, which constitute the majority of video data but are accessed infrequently. To address this challenge, this paper proposes Neural Cloud Storage (NCS), leveraging content-aware super-resolution (SR) powered by deep neural networks. By reducing the resolution of cold videos, NCS decreases file sizes while preserving perceptual quality. optimizing the cost trade-offs in multi-tiered disaggregated storage. This approach extends the cost-efficiency benefits to a greater range of cold videos and achieves up to a 21.2% reduction in total cost of ownership (TCO), providing a scalable, cost-effective solution for video storage.
Main Figure
