Joint Content Placement and Network Resource Allocation

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 290 papers121 conceptsfrozen 2026-10-04

In traditional network architecture, storage and transport are treated as independent silos. Content Delivery Networks (CDNs) decide where to replicate files, while Internet Service Providers (ISPs) decide how to route the resulting bits. This "separation of concerns" simplified early internet growth, but as data-intensive services like 4K streaming, AR/VR, and Large Language Model (LLM) inference scale, this decoupling has become a primary source of inefficiency.

A growing body of research argues for the joint optimization of content placement and network delivery. By treating storage (caching) and transport (routing, power, and bandwidth) as a unified resource pool, networks can proactively shape demand rather than reactively managing congestion.

The Coupling of Storage and Transport

The intuition behind joint optimization lies in a fundamental feedback loop: the location of a cache determines the traffic matrix of the network, and the state of the network determines the utility of a cache.

When these are optimized in isolation, the system often falls into suboptimal states. For instance, Weibo Chu and colleagues demonstrate that treating cache allocation and request routing as separate tasks leads to redundant content duplication and inefficient link usage. Conversely, Abhigyan Sharma and Ramesh K. Sitaraman show that in "Network-CDNs"—where ISPs operate their own caching infrastructure—sophisticated routing becomes less necessary. If you place content perfectly, the complexity of dynamic routing optimization yields diminishing returns because the traffic is already "shaped" at the source.

Cross-Layer Synergy

The benefits of this unified approach extend from the application layer down to the physical layer of wireless networks.

  • Physical Layer Interference: In 5G and 6G systems, cached content is not just stored data; it is a tool for signal processing. Khai Nguyen Doan and Mojtaba Vaezi explore how joint content placement and power allocation allow for better interference cancellation in Non-Orthogonal Multiple Access (NOMA) systems.
  • Radio Resources: Yan Kyaw Tun highlights that transmit power directly influences the effectiveness of edge storage. If a base station has the right content but lacks the power to reach a user efficiently, the cache hit is wasted.
  • Application-Aware Routing: In drone-based systems, Andrea Coletta and Gaia Maselli show that jointly tuning image compression and routing based on the specific task (e.g., object detection) significantly improves success rates compared to optimizing the network alone.

Multi-Tier Coordination

Modern networks are increasingly hierarchical, spanning from the cloud to the edge to the end-user's device. Joint optimization is critical for ensuring these tiers complement rather than compete with each other.

Chenchen Yang and Hao Feng demonstrate that jointly optimizing caching across dedicated infrastructure and peer-to-peer (D2D) devices maximizes the probability that a request is satisfied locally. Similarly, in the emerging field of satellite-terrestrial networks, Chang Liu and Yanmin Gong use Graph Neural Networks and quantum-assisted optimization to treat moving satellite constellations as proactive, distributed caches that mitigate terrestrial backhaul congestion.

Beyond Static Files: Model Caching

The paradigm is now shifting from caching static video files to caching computational models. As LLM inference becomes a dominant network load, Banghua Zhu and Lianmin Zheng have introduced frameworks that jointly optimize semantic caching (storing previous model outputs) and model multiplexing (routing queries to the most cost-effective model).

This represents the next frontier of the idea: the "content" being placed is no longer just a bitstream, but a mathematical sub-function. As shown by Shuting Qiu in the CoCaR framework, treating Deep Neural Networks as partitionable entities that can be partially cached and routed across the edge allows for high-precision inference even under strict latency constraints.

Conclusion

The research direction established by these authors suggests that the future of networking lies in breaking the abstraction between "where data lives" and "how data moves." Whether through DR-submodular optimization for congestion control or Deep Reinforcement Learning for hierarchical CDNs, the goal remains the same: a unified control plane that treats storage and transport as a single, dynamic resource.

Papers in this version

Showing 60 of 90 papers.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 75 papers
  2. Version 1 · 2026-10-04 · 90 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_joint_caching_and_routing_optimization_5f6f7bb0,
  title = {joint caching and routing optimization},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/joint-caching-and-routing-optimization/rdr_5f6f7bb0d3b14bd384f0b25ebd124dad},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_joint_content_placement_and_network_reso_5f6f7bb0_v1,
  title = {Joint Content Placement and Network Resource Allocation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/joint-content-placement-and-network-resource-allocation/rdr_5f6f7bb0d3b14bd384f0b25ebd124dad/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}