Future Internet Architectures: The Publish/Subscribe Internet Routing Paradigm

Alex Wanda
0
Since its conception, the Internet has experienced tremendous growth, ever increasing traffic and new applications, including voice and video, while still retaining its original architecture drafted almost 40 years ago. The main guiding principle for the design of the Internet was the end-to-end principle.

A number of challenges for the end-to-end Principle exist namely: operation in an untrustworthy Internet, more demanding applications, the rise of third party involvement, ISP service differentiation, and less sophisticated users.

Moreover, one of the most notable issues in the current Internet is the imbalance of powers in favor of the sender of information, who is overly trusted. The network accepts anything that the sender wants to send and will make a best effort to deliver it to the receiver. This has led to increasing problems with unsolicited traffic (e.g. spam e-mail) and distributed denial of service (DDoS) attacks.

The publish/subscribe (pub/sub) paradigm has been proposed as a remedy to the problems facing the current Internet. In pub/sub networking, senders “publish” what they want to send and receivers “subscribe” to the publications that they want to receive. In principle, no one receives any material to which they have not explicitly expressed an interest by way of subscription.

One can observe that many widely used Internet applications already are essentially publish/subscribe in nature. For example, distribution of software updates is currently performed in a poll/unicast fashion that is clearly non-optimal. Instead, subscribing to the updates that are needed and distributing them via multicast, caching etc. would be much easier and more efficient from the point of view of network resource usage. The PSIRP project will redesign the entire Internet architecture from the pub/sub point of view, taking nothing (not even IP) for granted. PSIRP’s work will focus on the intersection of security, routing, wireless access, architecture design, and network economics, in order to design and develop efficient and effective solutions. In such a new Internet, multicast and caching will replace unicast and cache-free data fetching operations, while security and mobility will be embodied directly into the foundation of the architecture rather than added as after-thoughts. The new pub/sub-based internetworking architecture aims to restore the balance of network economics incentives between the sender and the receiver and is well suited to meet the challenges of future information-centric applications and use modes. To our knowledge, this type of application of pub/sub communication models has not been tried before.

We should all aspire to change the routing and forwarding fabric of the global inter-network so as to operate entirely based on the notion of information (associated with a notion of labels to support fabric operation) and its surrounding concerns, explicitly defining the scope of the information and directly addressing information (via rendezvous identifiers) as opposed to addressing physical network endpoints. The envisioned operation on information is in sharp contrast to the current endpoint-centric networking model. The current end-to-end model of IP networking requires that both the relevant data and explicitly-addressed network locations be known in order to transparently stream information between two endpoints.

The PSIRP conceptual architecture is based on a modular and extensible core, called the PSIRP component wheel. The architecture does not have the traditional stack or layering of telecommunications systems, but rather components that may be decoupled in space, time, and context.

The novelty of the PSIRP proposal is to use publish/subscribe style interaction throughout the conceptual architecture, and thus support a layer-less and modular protocol organization. This organization is primarily achieved through the efficient structuring of information identifiers and their interactions amongst network elements, offering ample flexibility for future expansion.

The illustration below presents an outline of the conceptual architecture with the PSIRP component wheel in the middle. Above the wheel, we have APIs that facilitate accessibility to and implementation of different networking features that are available in the system. The figure illustrates the typical components needed in the wheel for inter-domain operation: forwarding, routing, rendezvous, and caching.

The global network of information can be viewed as an acyclic graph of related pieces of data, each identified and scoped by some identifiers. In the PSIRP architecture, identifiers define the relationships between the pieces of information in our system on the different levels, such as the application or networking level.



Post a Comment

0Comments

Post a Comment (0)
Subscribe Us

Get free daily email updates!

Follow us!