130 likes | 278 Views
end to end protocols for the Future Internet scope & goals. Bob Briscoe Chief Researcher, BT Group Dagstuhl, Jun 2008. why invite you?. from the theoretical end of practice and the practical end of theory
E N D
end to end protocols for the Future Internetscope & goals Bob BriscoeChief Researcher, BT GroupDagstuhl, Jun 2008
why invite you? • from the theoretical end of practice and the practical end of theory “The gap between theory and practice is greater in practice than in theory”Steve Crocker • people from distributed systems, data transport & internetworking fields; some cover both/all • mix of industry & academia • no attempt to be representative of ‘industry positions’ • expected to be able to take devils’ advocate positions • expected to question & argue • expected to desire timely consensus
goals • dialogue between fairly disconnected communities • (re)designing end-to-end data transport • (re)designing internetworking • reduce controversy over placement of functions • better consensus on layer interactions • both have to be clear for a robust system • produce prioritised research agenda • initial process towards architectural change (at end)
careful not to invent problems to fit the research we want to do research agenda from DARPA NewArch (2000) all still unsolved ‘solved’ = rough consensus and deployable code (ideally all solutions coherent) routing, naming, addressing (n) policy controls on inter-provider routing robustness & availability, inc mobility reachability through middleboxes resource control (0) highly time-variable resources capacity allocation extremely long propagation delays heterogeneity – cross-cutting agenda enabling conflicting socio-economic outcomes (0) enabling a variety of technical outcomes (n) management (0) policy-driven auto-configuration failure management security (n) attack resilience traceability you can’t have your dessertuntil you’ve eaten your vegetables (x) : no. of projects funded as ‘Future Internet’ (Spring 2007)
scope • future = changes with significant dependencies may take 5-10yrs to deployment so must last ~30yrs • Internet = globally interoperable platform for applications of computer communication • control & management focus more than data plane • routing, mobility, naming, addressing, interconnection • resource sharing, congestion control, framing/fragmentation • churn, rapid change of path characteristics • diversity of scale, technology diversity, availability, predictability • social & economic as well as technical control • not sexy but elegant – all pretty low level stuff
scoping &e2e terminology • recursion can be powerful or can confuse • end-to-end interaction in a distributed system / overlay network • end-to-end data communication • end-to-end of one internetwork hop over a network of links
...you should be interested in: • new communication paradigms and network architectures • cross-disciplinary motivations for new architecture – social, commercial and economic • applications enabled or improved through advanced networking and transport mechanisms • performance characteristics of new communication protocols and network architectures • availability and robustness aspects of new architectures • handling of unwanted traffic • transition and deployment aspects of new network architectures • formal principles of architecture and transport
out of scope • e2e issues not crossing traditional layer boundaries • information security • data integrity • multiplexing / demultiplexing • loss detection & repair • same order delivery • duplicate detection • flow control • specific link technology issues • application-specific networks
economic & social issuescan we stop others being myopic? • a different sort of research problem • if we agreed on a function placement ...say: • traffic engineering should be done by endpoints (role of networks is to give them the right incentives) • QoS should be done by endpoints (ditto) • policing should be done at outer trust boundaries • others would still pull control into their bailiwick • operators will still traffic engineer within their own network • the IEEE will still do QoS on each link • vendors will still put policing on each box
broken tension between e2e design principle and design for tussle • creating x-like systems out of un-x-like parts • where x is some desirable attribute • creating secure systems out of insecure parts • creating reliable systems out of unreliable parts • creating intelligent systems out of unintelligent parts • eg. intelligent session control without an intelligent network • creating QoS control systems out of non-QoS controllable parts • creating a video telephony system out of best effort Internet parts ... • creates low cost systems out of low cost parts with all the smarts at the ends... • creates profitable value chains out of unprofitable players...?
publicInternet open closed telco/NGN cellular satellite 1995 2008 open has not meant open minded • is the Internet only for those who want openness? • the worship of evolvability has to be balanced by a desire for the fruits of evolution to be enjoyed & exploited • it’s a tough research problem to enable both of: • intense competition & excess profits • distributed & centralised control • open & closed • designing for lock-in isn’t natural for most of us
end to end protocols for the Future Internetscope & goals Q&Aquestion & argue
goals – recap • dialogue between communities • placement of functions • layer interactions • prioritised research agenda • process towards architectural change