Chapter Summary

Chapter Summary

Key Points

• 1.

  Multi-server coded caching treats a network with $S$ cooperating transmitters, each with the library and $L$ antennas. Under full cooperation, the $S$ servers act as a single $SL$-antenna transmitter with joint precoding.

• 2.

  DoF formula: $\mathrm{DoF} = \min(t + S L, K)$. The caching gain $t$ remains unchanged; spatial gain multiplies by $S$. This extends Chapter 5's Lampiris-Caire result.

• 3.

  Shared vs dedicated caches. Dedicated: one cache per user, MAN baseline. Shared: $\Lambda$ caches across $K$ users, with $K_s = K/\Lambda$ users per cache. Rate per user scales linearly in $K_s$; at equal aggregate storage, shared can be more efficient when $K_s$ is moderate (fewer but larger caches).

• 4.

  The multi-server MAN scheme uses the standard MAN placement plus a cooperatively-precoded delivery: $(t + SL)$-subsets, $SL$ streams per subset. Per-server rate: $R_\text{MAN}/S$.

• 5.

  Cell-free massive MIMO + caching is the natural 6G deployment. Caches distributed across $S$ APs give aggregate gain $t = SM/N$; cooperation gives $SL$ spatial DoF. Total per-user DoF can reach near-unity in dense deployments.

• 6.

  Cooperation is expensive. Full cooperation requires low-latency inter-server backhaul. Practical deployments cluster 2-8 servers for full cooperation and time-share between clusters. The CommIT group's fog-mMIMO program prototypes these tradeoffs.

• 7.

  Distinguishing architectures. MAN user caching ($t = KM/N$) vs AP caching in CF-mMIMO ($t = SM/N$): the relevant aggregate cache differs. Don't confuse them.