A step toward practical deployment of the content-centric networking architecture

Yaogong Wang, Kyunghan Lee, Balakrishna Venkataraman, Ravi Lingappa Shamanna, Injong RheeStudentWorkshop@CoNEXT 2011 · DOI

TL;DR

This paper proposes a practical, cost-effective approach to deploying Content-Centric Networking (CCN) by leveraging existing Internet infrastructure. Instead of a complete overhaul, they introduce a hierarchical caching system and a novel Bloom filter-based mechanism for routers to efficiently advertise and discover cached content, aiming to reduce latency and redundant traffic.

What Problem Does This Paper Address?

The current Internet architecture, based on host-centric communication, faces challenges with efficiency, availability, and security. Content-Centric Networking (CCN) offers a promising alternative by shifting the focus from "who" (host) to "what" (content). In CCN, users request content by name, and any node can cache and serve it, potentially reducing latency, traffic, and server load. However, despite its theoretical benefits, practical deployment and implementation of CCN are scarce. Key challenges include efficiently advertising cached content, managing cache replacement, and enabling routers to make quick forwarding decisions in a scalable manner. This paper aims to address these practical deployment hurdles.

Methods

The authors propose a practical instantiation of CCN that builds upon existing Internet infrastructure, minimizing the cost of deployment. Their approach involves:

  • Hierarchical Caching: A three-level caching system is introduced:
  • Transient Cache: For local use, with immediate replacement.
  • Durable Cache: Contents cached for minutes, advertised to nearby routers.
  • Semi-Permanent Cache: Contents cached for hours, advertised across the Internet. Content popularity dictates its promotion or demotion between these levels.

  • Bloom Filter Generation and Aggregation: To inform other routers about cached content without overwhelming them with lists of names, Bloom filters (BFs) are generated.

  • Mitigating BF Size: Standard compression techniques are used for transmission. For local lookup, BFs can be "shrunk" by taking the union of their halves recursively.
  • Handling False Positives/Negatives: If the BF size is still an issue, they allow for high false positive rates (which are handled during forwarding) or stop insertion if the filter is nearly full, potentially leading to false negatives (which don't affect correctness).
  • FIB Construction: Advertised BFs are aggregated by routers. If advertisements arrive from the same interface, share the same next-hop, and are at the same distance, their BFs are combined using a union operation. This aggregated information forms a Forwarding Information Base (FIB) indexed by (interface, next-hop, hop count).
  • Forwarding Mechanism:
    • Linear Search of FIB: The FIB is designed to be small enough for a linear search. The first match found (sorted by hop count) indicates the nearest cache.
  • Handling No Match: If no match is found in the FIB, the request is forwarded via traditional IP after DNS resolution.
  • False Positive Handling: During forwarding, the probability of false positives in the matched BF is considered. If the product of hop count and false positive probability exceeds a threshold, the search continues, making the system more conservative for distant caches.
  • Pending Interest Table (PIT): The PIT mechanism from CCN is adopted to ensure data follows the request path.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Practicality over Purity: The core idea is that a "clean-slate" approach to CCN deployment is too costly. Leveraging existing infrastructure is a more viable path to initial adoption.
  • Hierarchical Caching for Scalability: The multi-level caching strategy intelligently manages content availability and advertisement scope based on popularity, balancing freshness and reachability.
  • Bloom Filters as a Scalable Advertisement Tool: BFs offer a compact way to represent sets of cached content, significantly reducing the overhead of advertising.
  • Intelligent False Positive Handling: The paper acknowledges the inherent trade-off in BFs (false positives) and proposes a mechanism to manage this risk during forwarding, making decisions based on the proximity of potential caches. This is a crucial step for practical CCN routing.
  • FIB Aggregation for Efficiency: Combining BF advertisements based on network topology (interface, next-hop, distance) creates a manageable FIB, avoiding the need for massive routing tables.

Key Results

The paper is an abstract and focuses on the proposed architecture rather than presenting quantitative results. The authors state that they will deliver more details and performance evaluation results during a poster session. The key qualitative outcome is the proposal of a system that aims to reap most CCN benefits with minimal cost and infrastructure overhaul.

Limitations and Future Work

  • Space Limitation: The authors explicitly state that due to space limitations in the abstract, many details are missing.
  • Performance Evaluation: Detailed performance evaluation results are not presented in this abstract.
  • Temporal Scalability: While the cache commitment mechanism is mentioned as addressing temporal scalability (update frequency), the details are not elaborated upon.
  • Spacial Scalability: The paper mentions addressing spacial scalability (routing table size) in Section 2.2, which is covered by the BF aggregation and FIB construction.
  • Future Work: The authors indicate that more details and performance evaluation results will be presented later, implying that further development and testing are planned.

Summary updated

BibTeX
@inproceedings{wang2011step,
  author = {Yaogong Wang and Kyunghan Lee and Balakrishna Venkataraman and Ravi Lingappa Shamanna and Injong Rhee},
  title = {{A step toward practical deployment of the content-centric networking architecture}},
  booktitle = {StudentWorkshop@CoNEXT},
  year = {2011},
  doi = {10.1145/2079327.2079347},
}
Figure