Full TitleFraMewoRk for a multI-ARCHitecture Internet
The success of the Internet is undeniable, as are its limitations. The most acknowledged architectural flaw is the lack of security in the original design, leaving the Internet vulnerable to various (increasingly common) attacks, while others question whether point-to-point packet delivery is still the appropriate service model in a content-oriented world. The research community responded to this problem with clean-slate redesigns of the Internet, including security approaches (e.g., SCION) and information-centric networks (e.g., NDN). So far, the impact of most of these clean-slate designs on the commercial Internet has been limited, for two main reasons. First, the Internet architecture is deeply embedded in its elements (that is, routers and end-host software). As a result, deployment is hard (the transition from IPv4 to IPv6 makes this point clear). Second, implementing a new Internet architecture that meets the expected performance is difficult. As networking hardware is fixed-function and the IP architecture is deeply entrenched, existing implementations are software-based, not fulfilling the performance requirements of any realistic network.
In this project, we propose to develop a framework, Myriarch, that enables a multi-architecture Internet that is both backward-compatible and enables evolvability. We, therefore, recognize the advances many architectures proposed over the past decade achieved in addressing specific limitations (e.g., content-centrality in NDN, security in SCION). But we advocate the need to go further, as a single architecture does not fit all needs. Myriarch will enable the inter-operation of multiple architectures, each evolving separately, leaving to applications the choice of the best for their needs. In addition, Myriarch will guarantee high performance, a crucial requirement for deployability. This will be achieved by building the translation mechanisms into programmable hardware (including programmable switching ASICs) and performant packet-processing software (NFV).
The design of Myriarch will build on ongoing collaborations and consult with world-leading experts in Internet Architecture (Jon Crowcroft, Univ. of Cambridge), security architectures (Adrian Perrig, ETH), NFV (Aurojit Panda, NYU), and Marie-Jose Montpetit (co-chair of IETF COINRG). The project’s advisory board, including Altice Labs, CAPGEMINI, and Telefonica Research will help us fine-tune the use cases and discuss potential deployment strategies.