[Paper Review] Efficient Multicasting in Content-Centric Networks Using Datagrams

J. J. Garcia‐Luna‐Aceves, Maziar Mirzazad Barijough|arXiv (Cornell University)|Aug 13, 2016

Caching and Content Delivery10 references3 citations

TL;DR

This paper proposes CCN-GRAM, a novel content-centric networking approach that enables native multicasting without maintaining per-Interest forwarding state using anonymous datagrams. By replacing traditional Pending Interest Tables (PITs) with Multicast Address Resolution Tables (MARTs), CCN-GRAM achieves similar or lower end-to-end delays while reducing forwarding table entries by over 150× compared to NDN, significantly improving scalability and resilience to flooding attacks.

ABSTRACT

The Named Data Networking (NDN) and Content-Centric Networking (CCNx) architectures are the leading approaches for content-centric networking, and both require using Interests (requests that elicit content) and maintaining per-Interest forwarding state in Pending Interest Tables (PIT) to store per-Interest forwarding state. To date, PITs have been assumed to be necessary to enable native support for multicasting in the data plane, such that multicast forwarding trees (MFT) are established by the forwarding and aggregation of Interests using PITs. We present a new approach to content-centric networks based on anonymous datagrams that provides native support for multicasting, but does so without the need to maintain per-Interest forwarding state. Simulation experiments are used to show that the proposed new approach attains the same end-to-end delays for multicasting while requiring orders of magnitude fewer forwarding entries.

Motivation & Objective

• To eliminate the need for per-Interest forwarding state (PITs) in content-centric networks while maintaining native multicasting support.
• To address the scalability and security limitations of PITs in NDN and CCNx, especially under high request rates or denial-of-service attacks.
• To design a data-plane mechanism that supports multicast forwarding without control-plane signaling or complex routing protocols.
• To evaluate the performance of the proposed approach under realistic multicast traffic conditions, focusing on forwarding state size and end-to-end delay.
• To enable efficient, scalable, and secure real-time content streaming in content-centric networks.

Proposed method

• Replaces traditional Pending Interest Tables (PITs) with Multicast Address Resolution Tables (MARTs) that store only one entry per multicast group, independent of request rate.
• Uses anonymous datagrams to carry Interest messages, eliminating the need to track individual Interests at intermediate routers.
• Employs a novel forwarding algorithm (Algorithm 3) that triggers multicast data transmission upon first receipt of a multicast Interest, even before all group members have signaled.
• Leverages reverse path forwarding (RPF) for data delivery, ensuring loop-free forwarding without per-Interest state.
• Maintains FIBs for name-based routing, while using MARTs to manage multicast group membership and forwarding state.
• Introduces a mechanism for efficient multicast tree construction and maintenance using only data-plane signaling, without control-plane protocols.

Experimental results

Research questions

• RQ1 Can native multicasting be supported in content-centric networks without maintaining per-Interest forwarding state?
• RQ2 How does the forwarding state size of CCN-GRAM compare to NDN under varying request rates and multicast group sizes?
• RQ3 What is the end-to-end delay performance of CCN-GRAM relative to NDN in multicast scenarios?
• RQ4 How does link delay affect the scalability and state growth of PITs in NDN versus MARTs in CCN-GRAM?
• RQ5 Can CCN-GRAM support high-throughput real-time streaming with lower latency and reduced resource consumption than NDN?

Key findings

• CCN-GRAM reduces the average number of forwarding entries per router by over 150 times compared to NDN under high request rates, with MART size remaining constant regardless of request volume.
• At 800 Interests per second and 15ms link delay, NDN’s average PIT size reaches 408 entries, while CCN-GRAM’s MART size remains at 5.26 entries on average.
• Under 30ms link delay, NDN’s average PIT size increases to 814 entries, with some routers storing up to 2,700 entries, while CCN-GRAM’s MART size remains unchanged.
• CCN-GRAM achieves lower end-to-end delays than NDN, especially in scenarios with asymmetric path delays, due to earlier multicast data transmission triggered by the first Interest.
• The average end-to-end delay in CCN-GRAM is consistently shorter than in NDN, even under high load, due to reduced forwarding state and more efficient tree formation.
• CCN-GRAM maintains scalability and reliability under high load and long delays, as MART size is independent of link delay or request rate, unlike PITs in NDN.