Problem
Research questions and friction points this paper is trying to address.
Two-layer Network Structure
Data Storage Optimization
Multilevel Caching Efficiency
Coded Caching for Hierarchical Two-Layer Networks with Coded Placement
🤖 AI Summary
Existing hierarchical caching schemes for two-tier networks (server → $K_1$ mirrors → $K_2$ users/mirrors) suffer from suboptimal bandwidth utilization due to separate, uncoded content placement across layers.
Method: We propose a hierarchical coded caching scheme with *joint coded placement*, explicitly optimizing the composite rate $ar{R} = R_1 + K_1 R_2$—a unified metric capturing total normalized backhaul ($R_1$) and mirror-to-user ($R_2$) bandwidth consumption—under global cache budget $ar{M}$. Our approach integrates coding-aware placement, hierarchical network modeling, and combinatorial design to enable cross-layer content reuse.
Contribution/Results: We formally characterize the fundamental trade-off between $ar{M}$ and $ar{R}$, and demonstrate that joint coded placement significantly reduces $ar{R}$ compared to conventional separation-based strategies under identical aggregate cache constraints. Both theoretical analysis and numerical experiments confirm the efficacy and superiority of our scheme in compressing multi-level bandwidth.
📝 Abstract
We consider the two-layered hierarchical coded caching problem introduced in [N. Karamchandani, $mathbf{U}$. Niesen, M. A. Maddah-Ali, and S. N. Diggavi, “Hierarchical coded caching,” IEEE Trans. Inf. Theory, 2016], in which a server is connected to $K_{1}$ mirrors, and each mirror is connected to $K_{2}$ users. The mirrors and the users are equipped with the cache of size $M_{1}$ and $M_{2}$, respectively. We propose a hierarchical coded caching scheme with coded placements that perform better than the existing schemes. In order to ensure a fair comparison with existing schemes, we introduce the notion of composite rate, defined as $overline{R}=R_{1}+K_{1}R_{2}$, which consists of the rate from server to mirrors $R_{1}$ and the rate from mirror to users $R_{2}$. The composite rate has not been discussed before in literature, and it represents the total consumed bandwidth in the system. Therefore, it is more appropriate to consider the composite rate along with $R_{1}$ and $R_{2}$. For the proposed scheme, we show a trade-off between the global memory $overline{M}=K_{1}M_{1}+K_{1}K_{2}M_{2}$ of the system and the composite rate. We compare the proposed scheme with the existing hierarchical coded caching schemes using the proposed parameter “composite rate.”
Innovation
Methods, ideas, or system contributions that make the work stand out.
Coded Caching
Composite Rate
Two-layer Network Structure
🔎 Similar Papers
No similar papers found.
