April 14 – 17 , 2009, Dagstuhl Perspectives Workshop 09162
Architecture and Design of the Future Internet
For support, please contact
This workshop brought together thirty seven experts from Europe, North America and Asia to discuss the way ahead for the Internet.
The aims of the workshop were as follows:
- to understand the 'state of the art' in Future Internet research by reviewing current programmes in the USA, Europe and Asia;
- to identify gaps and new opportunities in Future Internet research so that we can recommend new programme or project topics to appropriate bodies;
- to discuss potential collaborative activities amongst programmes or projects in order to make the most of current research: these activities could include testbeds and workshops.
During the workshop, the participants made contributions in:
- Defining and discussing the problem space. It was broadly agreed that three aspects are crucial: technological, economic, and societal/political;
- Describing relevant Future Internet activities in the USA, Europe and Asia;
- Remaining challenges in network essentials: naming/addressing, routing, mobility;
- How to provide for key network properties: security, resilience, performance;
- Management, policy, economic, green, and other Future Internet issues;
- Architecture questions: evolution vs revolution, virtualization, testbeds.
All of the above took place in plenary sessions, with a view to identifying the key issues that would be debated in smaller groups on the last day of the workshop. These key issues were as follows:
- Sacred cows;
- Management issues;
- Social, economic, green etc.;
- Programmability and virtualization;
- Personalization and context for Future Internet.
The participants self-organised into groups, which produced a summary of their discussion. Each group reported back in the final plenary session and a closing discussion followed.
Group (1) covered the IP address architecture, routing structure, TCP and the end-to-end argument, dumb core and smart edges – as the ‘sacred cows’ of the current Internet, and debated three things: layering principles, and whether there’s a need for management & control; re-routing as the primary approach to failure recovery, and whether overlays solve all problems; and virtual circuits (CO / CL).
Several of the comments indicated that we seem to be re-visiting these topics yet again, but perhaps in the light of new application or user needs (such as resilience) this is actually appropriate. Also, we don’t yet know best how to make the right choices from the above sets.
Group (2) offered some basic observations and issues that still require study: nested control loops, and stability provision; humans in the (control) loop – or not; knowledge, and the amount of data required, to produce satisfactory management – how to do inferences (we still don’t know how).
Group (3) covered a range of topics including the digital divide – which exists in all countries, the balance between security and privacy, network neutrality as a growing concern (or not), how to be ‘greener’ in networking, and the cultural and objective differences between academic and industry (for example ISPs) – where there will always be some tension.
Group (4) was concerned with whether programmability is now having its time, for example with the advent of multiple cores and the prospect of virtualization; associated research imperatives include router architecture, protocol architectures for massive parallelism, and the architecture of networks where the main routers have multiple cores.
Group (5) asked questions about personalization and context in the Future Internet: what is the typical usage; will virtual and physical worlds become more integrated, raising possible issues of privacy, social exchange etc.); which application areas will need further support, e.g. emergency, mobile video streaming, using social relationships to support communities, using public transport to support mobile users, etc.. Context was defined variously as what’s ‘around’, situation awareness, and it was agreed that context-modelling is an upcoming issue.
Unsurprisingly, in such a short time, the workshop participants did not manage to point definitively towards a new architecture or design for the Internet of the future. Rather, what the workshop did was to identify the beginnings of a number of promising tracks of investigation, listed above as (1-5), which are distinct from the activities currently being undertaken in the USA, Europe and Asia. These issues could form the basis of specifically-targeted seminars at Dagstuhl, and elsewhere, that have a much more interdisciplinary flavour and participant balance than the one reported here. It is clear that many diverse influences are being brought to bear on the possible Future Internet ‘shape’, and we should put together people from these diverse backgrounds in a focussed programme of discussions designed to elicit some more concrete outcomes. This would inevitably mean that the proportion of participating computer scientists and electrical engineers would have to be reduced considerably, balanced by a larger, carefully selected set of participants from other disciplines.
- Mobile Computing
- Future Internet architecture