Communications Research Challenges for the 21st Century David Hutchison, Saleem Bhatti, Ian Wakeman, Jon Crowcroft et al Executive Summary. We are not at the end of history. The existence of the cellular telephone network, digital radio TV and the Internet does not mean that the problems of communications have been solved. The reverse is the case. These networks are now built from and interconnect devices which are re-programmable. Thus the opportunities and risks are simultaneously increased exponentially compared with the analog, and fixed purpose digital networks of the previous century. We are critically dependant on an infrastructure, and yet it is being built on intuition, rather than scientific principles and sound engineering practice. It is not clear that one can retrospectively rectify this. It behoves us now to address the gaps in our knowledge, and learn whether these networks can safely evolve or be replaced by subsequent technologies based on sound understanding. The UK has great strengths in all three major types of networks at all levels, in industry and in academia, in services, technology, and in engineering, computing and content. yet we are losing ground to the far east and to some extent, with the US, despite leads in at least 2 of the major three approaches. In academic research, we have had a roller-coaster ride of funding in the community (which clearly exists, and the authors of this document believe they can represent at least partially). Since the development of the ideas in packet switching, the UK has been a world class player, with strong input to technologies and standards. Collaborations with BT, Nortel, with Vodafone, Orange, with BBC research, and with Cisco, Ericsson, Nokia, Sony etc etc abound. Yet we do not have a large single focused program of communications research in the UK academic world right now. We have a smorgasbord of programs, spread between various agencies (EPSRC, DTI, even PPARC, Industry, hybrids like CMI, Mobile VCE etc). The goal of this document is to spur the community to identify the programme of research work for the next 10 years that would lead to the networks that come after the Internet. The context is UK funding, and so the programme should speak to UK strengths, but also develop areas where we believe we should and could have strengths but do not at the moment. There are finite resources, and inevitably, therefore, there are areas that we should sunset. The existence of the BT 21CN and the UK Academic network are perhaps symptoms and signs of the direct areas to avoid. These represent mature networks that are built on the research results garnered over the 1980s and 1990s. To enable the UK networks of 2050 to exist, or indeed lead the world, the way these have done we need to have a coherent program of research, and now. Background Problems abound - denial and theft of service, unreliability, unpredictable costs and performance, lack of interworking (between broadcast, cellular and internet), lack of audit-ability. A common reaction to these problems is to close down those parts of the network design that appear to allow the "bad guys" in. In practice, this also closes down the possibility for innovation. The low barrier to entry in the Internet (and more recently in some mobile phone handset systems) has been the very thing that has created a lot of successful business, and yet is the reason that the Internet (and the cellular telephone net) are vulnerable, both to accidental and deliberate faults. As these systems become larger, more complex, and more critical, they become more vulnerable, at exactly the same time that we need to innovate precisely to mitigate the vulnerability. Let us consider what will happen when (eventually) the NHSnet spine, GPs and the public networks are interconnected (if you think security is a reason not to do this, then think again - it will be a legal requirement to provide data to the data owner - it wil lcertainly be a health risk to people if tey are not able to get at their own health records - since this is any member of the population, then necessarily , it will have to be disintermediated. This will be linkeed via medical telemmetry really soon (we have trial projects in this space already) and on to other networks (e.g. to bring in consultants to help paramedics at accident or crime scenes). Closed networks are a waste of public money, but open networks are a huge risk. Other examples abound. However, the community that worked in the last 25 years to bring these systems to fruition is at risk: a number of programmes of funding in the UK are coming to a close: the eScience program will naturally shrink as it changes leader at the DTI and the EPSRC CS related funding stream will no doubt slow This has until now funded some large, but rather specific communications work, mostly relating to Big Science use of fixed networks. To be fair, it has also largely been evolutionary rather than longer term revolutionary, or adventurous communications research, for very good reasons. the Equator IRC (and other IRCs) is nearing an end the Mobile VCE has largely gone into long ter industry-only mode ICT funding across the EPSRC is otherwise rather disjointed. since the end of the programmable networks programme BT and other similar labs are much reduced from their days of glory Only Microsoft research has significant presence in the UK, and communications is a minor interest of theirs. The only new program of work that is on an up is the Ubicomp stuff as exemplified by EPSRC Wines (1 and soon 2) calls - but this is more on the distributed systems side of the areas we work in which is very good, but we need the network underpinning, and there is no evidence for us that there is an end to the problems in sight. What we propose is to have a mini-UK challenge, similar to the larger one that the UKCRC and BCS sponsored, that led to some of the ubiquitous computing and other challenges being developed into EPSRC networks and programmes, but nowhere near as broad in scope. Strawman challenges. Some things we see as big bad problems out there to solve: linkage between network theory and practice in today's nets network science (network calculi, modern control theory, etc) safe extensible communications software understanding how to make explicit the compromises between security and monitoring. integrating and testing the mechanism design in protocol design complex systems (e.g. bio inspired distributed control) network resilience inter-provider routing, policy and QoS real working scaleable wireless mesh and ad hoc nets DoS detection and prevention worm and virus containment, use of trusted hardware delay&disruptive tolerant networking scaleable and efficient distributed management measurement and control reflective design to facilitate network monitoring mixed reality networks (e.g. road transport monitor and control) economics, policy and law of convergence (of mobile, fixed, data, voice and broadcast networks) linkage from underlay to distributed applications, both in practice, and in theory... Underpinning this is the increasing difficulty of doing practical experiments - the Ubicomp folks have attracted criticism in recent conferences for playing with mere toys - we in the underlay are in danger of not being able to work with real nets (UKERNA identified legal problems, constraints - also same problems in commercial ISPs), at least at a scale that is convincing....what is the right way forward? Perhaps because we have been tempted to look at ways to move forward from the existing Internet and wireless edges we have limited our vision. The document from the NSF workshop on "overcoming barriers to disruptive innovation in networking" lists some excellent proposals for similar activities in the USA. Some of the salient recommendations are relevant to the UK - their list includes: experimental systems research in networking validation of systems research in networking requirements for testbeds substantial increase in inter-disciplinary effort in the area creating synergy between different architectural approaches, rather than convergence help the community learn from industry Summary and conclusions The emergent BT 21CN, UMTS and digital broadcast networks all resulted from significant research in the UK in the 1980s and up until the 1990s successfully integrating knowledge gained into the commercial provider world. To position the UK to face the second half of the 21st century, research activity must be carried out now in the large, in the form of 'parallel' experimental network testbeds and programmes of study, for the next 20 years. This is essential for the continued survival (or indeed revival) of UK communications industry over the next 2 decades. The critical dependence of the economy, society, security, education and health on networks is well known. hat is less commonly acknowledged is how fragile (and brittle) these networks are. We need engineering and science principles as well as good intuition, to underpin networks of the future. Now is exactly the right time to embark on discovering these principles. References Overcoming Barriers to Disruptive Innovation in Networking Jin Turner et al, January 2005 http://www.arl.wustl.edu/netv/noBarriers_final_report.pdf The Communications Research Network, a CMI project http://www.communicationsresearch.net/about.html IAB Concerns and Recommendations Regarding Internet Research and Evolution http://www.faqs.org/rfcs/rfc3869.html