[Paper Review] LAC: Introducing Latency-Aware Caching in Information-Centric Networks - Technical Report

Giovanna Carofiglio, Léonce Mekinda+1 more|arXiv (Cornell University)|Jun 22, 2015

Caching and Content Delivery22 references3 citations

TL;DR

This paper proposes LAC, a latency-aware cache management policy for Information-Centric Networking (ICN) that dynamically prioritizes caching based on real-time, locally measured object retrieval latency. By storing objects with probability proportional to their latency, LAC reduces mean delivery time and standard deviation by up to 50% compared to LRU and probabilistic caching, while achieving faster convergence and lower network load.

ABSTRACT

Latency-minimization is recognized as one of the pillars of 5G network architecture design. Information-Centric Networking (ICN) appears a promising candidate technology for building an agile communication model that reduces latency through in-network caching. However, no proposal has developed so far latency-aware cache management mechanisms for ICN. In the paper, we investigate the role of latency awareness on data delivery performance in ICN and introduce LAC, a new simple, yet very effective, Latency-Aware Cache management policy. The designed mechanism leverages in a distributed fashion local latency observations to decide whether to store an object in a network cache. The farther the object, latency-wise, the more favorable the caching decision. By means of simulations, show that LAC outperforms state of the art proposals and results in a reduction of the content mean delivery time and standard deviation by up to 50%, along with a very fast convergence to these figures.

Motivation & Objective

• Address the lack of latency-aware cache management in ICN, despite its importance for 5G low-latency requirements.
• Overcome limitations of traditional cache policies (e.g., LRU, LFU) that ignore retrieval latency and network dynamics.
• Design a lightweight, distributed cache policy that uses local latency observations to guide insertion decisions without complex coordination.
• Improve end-to-end delivery performance by prioritizing long-latency content for caching, thereby reducing overall retrieval delay.
• Ensure fast convergence and stability in dynamic environments, outperforming probabilistic caching approaches with similar complexity.

Proposed method

• Leverage locally measured retrieval latency—the time between request forwarding and packet reception—as the primary metric for cache insertion decisions.
• Define a cache insertion probability proportional to the observed latency of each object, favoring those that take longer to retrieve.
• Integrate LAC as an add-on to any existing cache replacement policy (e.g., LRU), modifying only the insertion criterion.
• Use a dynamic function f that tracks the mean latency of all cached objects to adapt the insertion probability over time.
• Implement the policy in a decentralized manner, requiring no global knowledge or protocol modifications.
• Apply the mechanism in a probabilistic fashion: higher latency objects are more likely to be cached, even if less popular.

Experimental results

Research questions

• RQ1 Can latency-aware cache insertion significantly reduce mean and variance of content delivery time in ICN?
• RQ2 How does LAC compare to classical LRU and probabilistic caching (e.g., LRU+LCP) in terms of convergence speed and delivery performance?
• RQ3 Does using local latency observations as a proxy for network distance or congestion improve caching efficiency without increasing complexity?
• RQ4 Can a measurement-based, low-complexity policy like LAC achieve better performance than latency-insensitive policies in dynamic ICN environments?
• RQ5 What is the impact of LAC on network load distribution and link utilization across different topologies?

Key findings

• LAC reduces mean content delivery time by up to 50% compared to standard LRU in both line and tree network topologies.
• LAC reduces the standard deviation of delivery time by up to 50%, leading to more stable and predictable user experience.
• In a line topology, LAC achieves a 50% reduction in mean delivery time and outperforms LRU+LCP in convergence speed and stability.
• In a tree topology, LAC reduces mean delivery time by 30% and lowers maximum and average link load compared to LRU and LRU+LCP.
• LAC significantly reduces the overall cache miss probability and stabilizes delivery time, especially under high variability in content popularity.
• LAC converges faster than LRU+LCP and exhibits lower standard deviation in delivery time, indicating superior dynamic adaptability.