Multicast forwarding in future Information-Centric network architectures

Abstract

Information-Centric Networking (ICN), an active research area in Computer Networking, has the goal of designing a network architecture that better suits today's networked world: billions of connected devices requesting access to voluminous amounts of digital media. ICN architectures aim to facilitate content distribution by placing self-identified information items at the heart of the networking protocol stack and building routing and transport protocols around them. Among other design choices, ICN emphasizes support for scalable and efficient multicast delivery. However, practical implementations of ICN architectures have found that achieving this goal is easier said than done. For instance, the Content-Centric Networking (CCN) architecture realizes single-source multicast delivery through a distributed stateful forwarding scheme, which faces scalability issues with respect to the forwarding state kept at routers. On the other hand, the Publish-Subscribe Internet (PSI) architecture proposes a centrally-controlled fully stateless multicast forwarding scheme with In-packet Bloom filters (IBF), which faces scalability issues with respect to the network and/or multicast group size. This dissertation addresses the issue of multicast forwarding scalability in two very different ICN architectures, namely CCN and PSI. The common ground in the proposed solutions is the use and extension of Bloom filter-based packet forwarding mechanisms. In CCN, we relax the fully stateful nature of its forwarding scheme, in order to improve the architecture's scalability with respect to the forwarding state kept at routers. We propose a semi-stateless forwarding scheme by incorporating Bloom filter-based forwarding, while respecting the architecture's distributed routing and forwarding operations. Our simulation-based study shows that the scheme can effectively reduce forwarding state at routers by 60%-70% at the cost of 1.5%-12% in bandwidth overhead. In PSI, we similarly relax the fully stateless nature of its forwarding scheme, proposing a semi-stateless alternative with IBF switching, in order to allow IBFs to efficiently scale to any network and/or multicast group size. We integrate IBF switching in PSI, utilizing the architecture's centralized control-plane operations. Our simulation-based study shows that with IBF switching, forwarding scalability is achieved regardless of network and/or multicast group size, at the cost of placing forwarding state at 0.5%-2.5% of forwarding nodes, which is far less than that of competing technologies.