Staying focused on the big unsolved problems e.g. resource accountability Bob Briscoe Chief Researcher, BT Group Sep 2008

2 main problems with the current Internet interconnecting the information world with the physical world publish-subscribe commercial viability & flexible comms industry structure resource accountability policy controls on overlay / inter-provider routing scaling for extreme dynamics mobility, re-routes & failures, time-variable resources high speed transfers, very short transfers availability & robustness to attack (not infoSec problem, nec.) malware transmission & DDoS reachability for the good guys through middlewalls management failure tracking policy-driven auto-config

3 research style – keeping focused all need a 'big science' approach monumental but tedious research process sustainable research teams consensus on what each problem is separately –based on rigorous analysis (engineering & economic) a range of solutions proven much better than today for each problem –rigorously analysed and strong economic cases for deployment process for rough consensus on solution(s) to deploy computing / networking research not set up like that the next sexy start-up idea from the lone PhD research as marketing – crafted to hide weaknesses little critical analysis of other work emphasising novelty de-emphasising commonality of analytical model specialisation, not abstraction "Hey, if you try this instead of that, simulations show it rocks and it's much better." "We've invented this cool way to share resources in an ad hoc, cognitive, autonomic, meshed network of tetherless hospital beds, which is probably novel based on a survey of papers found on the first page of a Google search."

4 Flow-Rate Fairness: Dismantling a Religion experience of trying to bring science into Internet research 2Mbps access each 80 users of attended apps 20 users of unattended apps usage typeno. of users activity factor ave.simul flows /user TCP bit rate /user vol/day (16hr) /user traffic intensity /user attended805%=417kbps150MB21kbps unattended20100%=417kbps3000MB417kbps x1x20 time flow activity rate 10Mbps

5 two arbitrary approaches fighting bit-rate time 'flow-rate equality' throttling heavy volume usage the Internet way (TCP)operators (& users) degree of freedom‘flow rate equality’‘volume accounting’ multiple flows  activity factor  application control  congestion variation  each arbitrarily cancels out the worst failings of the other Internet looks like 'it works OK' but arbitrary fighting leaves collateral damage

6 fairer is faster – incentivise end host behaviour bit-rate time light heavy light heavy light heavy 'unfair' TCP sharing heavier usage takes higher sharing weight throttling heavy usage enabler: limit congestion, not volume then end system congestion control will quickly evolve heavy usage will back away whenever light usage appears so light usage can go much faster hardly affecting completion times of heavy usage differentiated QoS as if in the network lighter usage takes higher sharing weight

7 limiting congestion with flat fee pricing only throttles traffic when contribution to congestion elsewhere exceeds allowance otherwise free to go at any bit-rate bulk congestion policer congestion · bit-rate 0% · 2 Mb/s = 0.0kb/s 0.3% · 0.3Mb/s = 0.9kb/s 0.1% · 6 Mb/s = 6.0kb/s 6.9kb/s Internet 0.3% congestion 0% 0.1% 2 Mb/s 0.3Mb/s 6 Mb/s Acceptable Use Policy Your 'congestion volume' allowance: 1GB/month (= 3kb/s continuous) This only limits the traffic you can try to transfer above the maximum the Internet can take when it is congested. Under typical conditions this will allow you to transfer about 70GB per day. If you use software that seeks out uncongested times and routes, you will be able to transfer a lot more. Your bit-rate is otherwise unlimited

8 mismatch of aspirations experience of trying to bring science into Internet research [y] Kelly's "Charging & Rate Control for Elastic Traffic": 1997 [z] "Dismantling a Religion" repeated the message: 2006 a rich and rigorous vein to tap for fixing the Internet resource accountability, simpler e2e QoS mechanism, e2e QoS for mobility, differentiated flow start-up, interconnection contracts, targeted liability for DDoS, traffic engineering and congestion routing, removing the need for deep packet inspection, removing need for rate limits into shared access networks (PON, cable, wireless)... context of all citations so far (paraphrasing) "There are a number of approaches to fairness [w,x]; recently these have been deeply challenged [y|z]; but we build on [x] which is generally used in the research community." Typical mailing list discussions "What's the problem? TCP works" ug

Staying focused on the big unsolved problems e.g. resource accountability Q&A