[Paper Review] Accounting for Information Freshness in Scheduling of Content Caching

Ghafour Ahani, Di Yuan|arXiv (Cornell University)|Oct 29, 2019

Age of Information Optimization18 references4 citations

TL;DR

This paper proposes a joint optimization framework for dynamic content caching in base stations, balancing backhaul load reduction and content freshness via age of information (AoI). It formulates the problem as an NP-hard integer linear program, solves it using column generation with a tailored rounding heuristic, and demonstrates near-optimal performance within 1% of the lower bound in simulations.

ABSTRACT

In this paper, we study the problem of optimal scheduling of content placement along time in a base station with limited cache capacity, taking into account jointly the offloading effect and freshness of information. We model offloading based on popularity in terms of the number of requests and information freshness based on the notion of age of information (AoI). The objective is to reduce the load of backhaul links as well as the AoI of contents in the cache via a joint cost function. For the resulting optimization problem, we prove its hardness via a reduction from the Partition problem. Next, via a mathematical reformulation, we derive a solution approach based on column generation and a tailored rounding mechanism. Finally, we provide performance evaluation results showing that our algorithm provides near-optimal solutions.

Motivation & Objective

• Address the trade-off between backhaul load reduction and content freshness in time-varying content caching environments.
• Formulate a joint optimization problem that accounts for both content popularity (request frequency) and age of information (AoI).
• Develop a solution framework that respects limited cache capacity while minimizing a composite cost function of downloading, updating, and AoI penalties.
• Prove the NP-hardness of the caching scheduling problem via reduction from the Partition problem.
• Design a scalable solution approach that provides near-optimal performance for practical deployment in edge networks.

Proposed method

• Formulate the caching scheduling problem as an integer linear program (ILP) with time-varying content popularity and AoI-based freshness constraints.
• Reformulate the ILP using Dantzig-Wolfe decomposition to enable column generation, transforming the master problem into a set partitioning form.
• Model the pricing subproblem as a shortest path problem in a time-expanded network, solvable in polynomial time.
• Introduce a problem-specific rounding algorithm (RA) to convert fractional solutions from column generation into feasible integer cache schedules.
• Use the column generation algorithm to compute a tight lower bound (LB) on the global optimum, enabling performance evaluation.
• Integrate the cost model where downloading, updating, and AoI are weighted via a penalty function with parameter λ.

Experimental results

Research questions

• RQ1 How can content caching scheduling be optimized to jointly minimize backhaul load and age of information (AoI) under time-varying popularity?
• RQ2 What is the computational complexity of the joint caching and freshness optimization problem?
• RQ3 Can a column generation-based approach with tailored rounding yield near-optimal solutions for this NP-hard problem?
• RQ4 How does the proposed algorithm compare to conventional caching strategies (e.g., random or popularity-based) in terms of cost, AoI, and backhaul load?
• RQ5 What is the impact of system parameters such as user count, file set size, and cache capacity on the performance of the proposed scheme?

Key findings

• The proposed column generation algorithm (CGA) achieves solutions within 1% of the lower bound, indicating near-optimality.
• The optimality gap decreases with increasing user count (U), as higher request diversity makes optimal caching more predictable.
• Cost is minimized at moderate file set sizes (F ≈ 150), where cache capacity scales with file diversity without overwhelming the system.
• When λ increases (favoring lower AoI), conventional algorithms (PBA, RBA) drastically increase backhaul load, while CGA maintains a balanced trade-off.
• The column generation subproblem is reduced to a shortest path problem, enabling polynomial-time solution and efficient convergence.
• The tailored rounding algorithm (RA) effectively transforms fractional solutions into feasible integer cache schedules with minimal performance loss.