[Paper Review] Efficient Proactive Caching for Supporting Seamless Mobility

Vasilios A. Siris, Xenofon Vasilakos+1 more|arXiv (Cornell University)|Apr 18, 2014

Caching and Content Delivery10 references4 citations

TL;DR

This paper proposes a distributed, congestion pricing-based proactive caching scheme that uses mobility predictions to pre-cache data in proximity to future mobile attachment points, reducing transfer delay. It achieves robust performance close to optimal and oracle schemes, even under varying mobility patterns and cache constraints, by efficiently allocating limited storage using dynamic pricing.

ABSTRACT

We present a distributed proactive caching approach that exploits user mobility information to decide where to proactively cache data to support seamless mobility, while efficiently utilizing cache storage using a congestion pricing scheme. The proposed approach is applicable to the case where objects have different sizes and to a two-level cache hierarchy, for both of which the proactive caching problem is hard. Additionally, our modeling framework considers the case where the delay is independent of the requested data object size and the case where the delay is a function of the object size. Our evaluation results show how various system parameters influence the delay gains of the proposed approach, which achieves robust and good performance relative to an oracle and an optimal scheme for a flat cache structure.

Motivation & Objective

• To reduce data transfer delay in mobile networks by proactively caching data closer to future user attachment points using mobility predictions.
• To efficiently utilize limited cache storage in distributed, multi-level cache hierarchies through a congestion pricing mechanism.
• To address the hard problem of proactive caching with variable-sized objects and hierarchical cache structures.
• To evaluate the scheme's robustness to uncertainties in mobility patterns and object size-dependent delays.
• To enable distributed decision-making for caching without requiring global coordination or object popularity data.

Proposed method

• The scheme uses mobility prediction to estimate the probability of a mobile connecting to future network points, guiding where to cache data.
• Congestion pricing dynamically allocates cache space based on demand and delay gains, ensuring high-value objects are cached first.
• The approach models both fixed and variable delay scenarios—where delay is independent of or dependent on object size.
• It supports flat and two-level cache hierarchies, allowing mid-level and leaf-level caching decisions based on mobility and delay trade-offs.
• Pricing updates are performed iteratively using a price update factor γ, enabling convergence to near-optimal caching decisions.
• The system operates in a distributed manner, with each cache making independent decisions based on local mobility and pricing information.

Experimental results

Research questions

• RQ1 How does proactive caching based on mobility prediction reduce data transfer delay in mobile networks?
• RQ2 Can congestion pricing effectively manage limited cache capacity in a distributed, multi-level caching architecture?
• RQ3 How does the scheme perform under varying mobility patterns and object size-dependent delays?
• RQ4 What is the impact of uncertainty in mobility transition probabilities on the scheme’s performance and convergence?
• RQ5 How does the scheme compare to an oracle and optimal schemes in terms of delay reduction and robustness?

Key findings

• The proposed EPC scheme achieves over 95% of the steady-state delay gain within approximately 490 handovers when the price update factor γ is 8, indicating fast convergence.
• For smaller γ (e.g., 0.5), convergence takes longer (~750 handovers), showing that slower price updates delay adaptation to mobility changes.
• The scheme’s delay gains are robust to variations in mobile transition probabilities: a 30% standard deviation in probabilities reduces gains by less than 10%.
• The oracle scheme is more sensitive to high transition probability variation, as it is forced to use mid-level caches with lower delay gains when leaf-level caching is insufficient.
• In a scaled-down Internet topology, the oracle’s performance degrades more than EPC’s due to increased topological variation, while EPC maintains strong performance.
• The EPC scheme outperforms naive caching and approaches the performance of an optimal scheme for flat cache structures, even under complex, real-world conditions.