Impacts of Device Caching of Content Fractions on Expected Content Quality

Dongjae Kim, Minseok Choi

Abstract

Traditional wireless caching strategies typically store entire video files, which can lead to inefficient storage use and rigid quality levels. This paper proposes a novel approach: caching high-quality fractions of multiple popular videos. By optimizing the specific chunks and quality levels stored, the authors demonstrate significant improvements in expected video quality, particularly in high-connectivity environments.

Introduction

As mobile video traffic continues to surge, edge caching has emerged as a critical technique to alleviate network congestion and reduce latency. However, conventional models often treat video content as monolithic blocks, leading to two primary issues: storage bias, where large files dominate limited device memory, and quality rigidity, where users are forced into binary "cached or not" scenarios regardless of available network bandwidth.

This research addresses these challenges by formulating an optimal caching policy that determines what fraction of a video to store and at what quality level to maximize the "Expected Content Quality" delivered to the user.

Methodology

The study models a wireless network where a Base Station (BS) serves devices with finite storage capacities. Video content is treated as divisible into chunks that can be encoded at varying quality levels.

A central finding of the mathematical analysis is that it is always optimal to cache the highest possible quality ($q_{max}$) for a fraction of the content, rather than a larger portion at a lower quality. To implement this, the authors developed algorithms based on "water-filling" and bisection search techniques to solve the optimization problem for both fixed-distance and random-distribution scenarios.

The model incorporates Rayleigh fading and path loss, ensuring the BS effectively "fills in" the uncached portions of the video based on current channel conditions.

Key Results

The proposed fractional caching policy consistently outperformed traditional "entire-content" caching across all tested spatial distributions.

Impact of Connectivity

The performance advantage of fractional caching is most pronounced in high-connectivity environments. As the transmit Signal-to-Noise Ratio (SNR) increases, the gap between the proposed method and traditional caching widens. In optimal channel conditions, the expected quality approaches the theoretical maximum ($q_{max}$), as shown in Figure 4.

Expected quality vs. SNR for various caching and distribution scenarios

Distance Sensitivity

While fractional caching offers clear benefits, its effectiveness is sensitive to the device's distance from the Base Station. As the average distance $E[r]$ increases, the network's ability to reliably deliver the uncached portions of the video diminishes, leading to a decrease in expected quality, as illustrated in Figure 5.

Impact of average device distance on expected content quality

Discussion and Future Work

The core insight of this work is that "quality trumps quantity" in device caching. By storing high-bitrate foundations of many popular videos, devices can act as quality boosters rather than simple storage bins. This approach also promotes fairness and diversity, preventing a few top-ranked videos from monopolizing cache space.

Future research directions include adapting the model for discrete quality levels (as used in DASH/HLS protocols) and exploring Device-to-Device (D2D) assisted networks. In D2D scenarios, users could potentially fetch different fractions of a video from multiple nearby peers, further enhancing network efficiency and user experience.

Summary updated

BibTeX
@article{kim2022impacts,
  author = {Dongjae Kim and Minseok Choi},
  title = {{Impacts of Device Caching of Content Fractions on Expected Content Quality}},
  journal = {IEEE Wireless Communications Letters},
  year = {2022},
  doi = {10.1109/lwc.2022.3153125},
}
Figure