Chapter Summary

Key Points

1.
Video is the killer app for coded caching. 80% of internet traffic is video; caching that helps video is transformative.
2.
Multi-rate coded caching handles successively-refinable content: rate $= b_{\max} \cdot R_\text{MAN}$ , preserving the $1 + K\mu$ gain at each quality level.
3.
Optimal cache split across SR layers: $\alpha_\ell^* \propto p_\ell b_\ell$ (popularity × bitrate). Convex optimization — tractable and unique.
4.
CommIT contribution (Zhang-Moharrami-Caire 2020) establishes the multi-rate coded caching framework, integrating rate- distortion theory with multicasting.
5.
QoE is the practical metric. Composite of bitrate, startup latency, rebuffering, smoothness. Coded caching helps all four; rebuffering reduction dominates user-perceived gain.
6.
Chunked delivery is the practical unit: 2-10 second segments. MAN applies at chunk granularity; subpacketization is more practical due to chunk size.
7.
DASH / HAS integration. Full MAN gain requires MBMS-style multicast or HTTP/3 server push. Opportunistic $\sim 50\%$ gain over standard DASH. 5G MBMS is first commercial venue.
8.
Deployment roadmap. 5G MBMS scalable video: 2025-2027. CMAF harmonization: ongoing. HTTP/3 server push: 2026+. Commercial coded-DASH: 2026-2028 target.

Looking Ahead

Chapter 22 closes the book with open problems: the gap between uncoded-placement and coded-placement optimality, heterogeneous caches, finite-blocklength effects, and integration with emerging network architectures. It's a natural coda to the 22-chapter tour: what we know, what we don't, and where the next decade of research is headed.

HTTP Adaptive Streaming (HAS / DASH)Exercises