I’ve been involved in some research recently which has more than a passing interest in the networking concepts of RINA a clean slate Internet architecture which is proposing to resolve the challenges of today’s internet.
The reason for the interest is that RINA has brought the networking challenge back to it’s basics, and bases itself on the concept that networking is just Inter Process Communication (IPC).
This is to say in a RINA world any two application processes in different systems are able to communicate using the services provided by a DIF. A DIF is nothing more than a number of cooperating application processes, with the DIF as the structural unit for organizational purposes, which in another way could be referred to as a ”layer”.
A DIF is different from the traditional definition of layer in the Transmission Control Protocol/Internet Protocol (TCP/IP) architecture. Firstly, a DIF does not perform a single function or small subset of pre-determined functions, but a coordinated set of policy managed functions to achieve the desired IPC service.
Secondly, the DIF naturally separates various concerns, including operation over different timescales (e.g. short -term data transfer and multiplexing vs. long-term connection management and access control issues). By the way DIFs are repeated, making RINA recursive.
This proposal of a new network architecture through RINA is currently being evaluated by its ability to address the shortcomings of the current Internet architecture and how it manages to provide solutions to these issues.
To paraphrase the researchers at Boston University (BU), RINA tackles the following set of challenges and provides an answer to most of them without extra effort but inherently:
Multihoming: By adopting and extending Saltzer’s proposal [1] for a naming and addressing schema, RINA names nodes as well as interfaces (Point of Attachment (PoA)). Thus, it is able to identify a node by its name and achieve multihoming.
Mobility: RINA simply sees this as a dynamic version of multihoming with controlled “link failures”, i.e., as a wireless signal becomes weak, the link “fails”.
Multicast: For RINA all addressing (anycast, multicast) can be treated as a set of addresses and a rule. The rule determines the number of members in a set that satisfy the rule.
Security: RINA addresses security in that a DIF provides a secure container. Users of the DIF only see the destination application name and a local handle. RINA does not use addresses nor well- known ports.
Policy Based Configuration: With RINA, policy and mechanism are separated. By using policies in conjunction with the common mechanisms, RINA can be configured to meet the different requirements of applications.
A detailed overview of RINA innovations and features can be found in the papers [2, 3, 4]. Tutorials and reference material are available from the Pouzin Society (PSOC) website.
[1]. J. H. Saltzer. On the naming and binding of network destinations. In Proceedings IFIP/TC6 International Symposium on Local Computer Networks, pages 311–317, April 19-21 1982.
[2] J. Day. How PNA Works: The Future of Networking [pdf].
[3] J. Day. Patterns in Network Architecture: A Return to Fundamentals (ISBN 0132252422), December 2007.
[3] J. Day, I. Matta, and K. Mattar. “Networking is IPC”: A guiding principle to a better internet [pdf]. In CoNEXT ’08: Proceedings of the 2008 ACM CoNEXT, pages 1–6. ACM, 2008.