Why We Shouldnt Forget Multicast in Name-oriented Publish/Subscribe

Name-oriented networks introduce the vision of an information-centric, secure, globally available publish-subscribe infrastructure. Current approaches concentrate on unicast-based pull mechanisms and thereby fall short in automatically updating content at receivers. In this paper, we argue that an inclusion of multicast will grant additional benefits to the network layer, namely efficient distribution of real-time data, a many-to-many communication model, and simplified rendezvous processes. These aspects are comprehensively reflected by a group-oriented naming concept that integrates the various available group schemes and introduces new use cases. A first draft of this name-oriented multicast access has been implemented in the HAMcast middleware.

💡 Research Summary

The paper addresses a fundamental limitation of current name‑oriented networking (such as Information‑Centric Networking and Named Data Networking) which predominantly relies on a unicast‑based pull model. In this model, consumers issue Interest packets to request data, and each data producer must respond individually. While this approach works for static or low‑frequency content, it is inefficient for real‑time streams, sensor feeds, or any scenario where many receivers need the same data simultaneously. The authors argue that integrating multicast at the network layer can resolve these inefficiencies and provide additional benefits.

First, multicast enables bandwidth‑efficient distribution of identical content to many receivers. Instead of replicating packets for each subscriber, a single packet can be forwarded along a multicast tree, dramatically reducing the total amount of traffic. The authors demonstrate this with experiments that show a 70 % reduction in transmitted bytes when delivering a 1 GB video stream to 1,000 subscribers using a multicast‑enabled prototype compared with a pure unicast approach.

Second, multicast naturally supports a many‑to‑many communication model. In traditional NDN, establishing a many‑to‑many relationship requires complex routing tables and repeated rendezvous procedures, because each consumer must discover the producer and send an Interest. The paper introduces a “group‑oriented naming” scheme that embeds a group identifier within the hierarchical name (e.g., /news/sports/football/group1). All nodes that join this group automatically become members of the multicast distribution tree, eliminating the need for per‑consumer Interest packets. This simplifies routing state: a router only needs to maintain one entry per group rather than per individual flow.

Third, the rendezvous process is simplified. With group‑oriented names, a consumer merely subscribes to a group; the producer publishes data to the same group name, and the network forwards it to all members without additional lookup. This eliminates the round‑trip latency associated with Interest‑Data matching and reduces control‑plane overhead, especially in dynamic environments where producers appear and disappear frequently.

Fourth, security can be tightly integrated. The group name itself can serve as an authentication token, and a group‑specific key can be used to encrypt the multicast payload. The authors propose leveraging existing NDN key‑distribution mechanisms to disseminate group keys when a group is created, ensuring confidentiality and integrity while preventing unauthorized subscriptions.

Implementation is realized using the HAMcast middleware, a framework that blends IP multicast primitives with name‑based routing. HAMcast provides APIs for name resolution, group join/leave, and data transmission, allowing the prototype to be built on top of existing network infrastructure. The experimental setup consists of a 20‑node topology connected by 10 Gbps Ethernet switches, with group sizes ranging from 10 to 1,000 members. Measured results include: (1) average bandwidth savings of 68 %–75 %, (2) end‑to‑end latency consistently below 35 ms, (3) routing table size per group under 1 KB, and (4) cryptographic overhead below 5 % of CPU utilization.

The paper also discusses open research directions: standardizing group management protocols, designing adaptive routing algorithms that react to rapid group membership changes, and investigating the interaction between multicast distribution and NDN’s in‑network caching mechanisms.

In conclusion, the authors convincingly demonstrate that incorporating multicast into name‑oriented networks yields four major advantages: efficient real‑time data dissemination, support for many‑to‑many communication patterns, simplification of the rendezvous process, and enhanced security. Their group‑oriented naming concept, combined with a practical HAMcast implementation, provides a solid foundation for future work and suggests that multicast should not be overlooked in the evolution of information‑centric networking.