TF-MSP / TF-PR Z ü rich 30 September 2008 John DYER TERENA John.Dyer@terena.org Developing The Case for NRENs ( A BI T MORE) revised 0 8 -October-2 0 0 8
W here did w e get up to since 1 8 May 2 0 0 8 ? › DRAFT for DI SCUSSI ON version 1 › www.terena.org/ activities/ tf-msp/ documents/ nren-case-v1.pdf › Suggestions at last m eeting › Presentations in TNC 2 0 0 8 › Discussions at the GA › Em ails on the TF-MSP list › Com pendium data and trends › Discussions w ith CEO of REANNZ Slide 2
Major Suggestions from May TF-MSP/ PR m eeting › Different NRENs have different situations › Create a num ber of scenarios W hat-if: I ssues w ith Regulator › Commercial / Competition issues Dissatisfaction from the user/bill-payer Lack of Political Support › There are potential dangers in the environm ent in w hich w e operate › Keep aw are of regulatory, political & com m ercial im pact of our portfolios m ay have Slide 3
Presentations during TNC 2 0 0 8 › som e serious questions about the future of research netw orks. › Do NRENs need to develop m ore functionality? › Should NRENs think about a new business m odel? › Should NRENs rem ain separate from other ( public) services? › I f NRENs do, w ill they die, be superseded by the m ore rapidly developing com m ercial sector, or continue alongside as a niche m arket? › have to offer w hat people w ant, not necessarily the technology that is best Slide 4
TERENA GA Discussions May 2 0 0 8 › “ FREE” services - being used by NREN som e custom ers Are they really FREE? W hat are the costs, im plications? › NRENs should m ake use of their position and explore new › opportunities › I ncreasing NREN collaboration on Cross-Border-Fibres Relies less on centralised international connectivity m odel › Requires com m on agreem ents on SLA, CP, AUP, Security › › Connections becom ing available at prices below those currently being paid in the NREN com m unity Procurem ent by NREN at national level is cost effective › NRENs are able to provide services tailored to the com m unity › Users value the services › End-to-End com m unity can sort out issues ( PERT) › Slide 5
Em ail Discussions › We are here only to foster tele-informatic services in higher education and research › We found building a community is useful › Whenever services become mainstream pull-out › NRENs should be better and cheaper than the market? or › As the gap between ISP offering and NREN services closes in terms of price and capability NRENs should: a) compete on equal terms ? b) disappear ? c) re-think their role ?
Com pendium 2 0 0 8 findings › NREN Traffic The NREN approach to QoS › W here is the traffic going › I Pv6 rollout › › Funding Econom ic Models › › Agency/ Principal v Transaction Costs › Free m arket › Leading to the conclusion that › Com petition is better than Cooperation ? › Hybrid Solution ? Slide 7
NREN Traffic to External Netw orks % External Link Utilisation %T3 and %T4 in 2008, EU/EFTA 25% � 20% � � 15% � %T3 %T4 � � 10% � 5% 0% Moldova Slovakia Hungary Czech Republic Switzerland Latvia Malta Germany Italy Morocco Ireland United Kingdom Netherlands Israel Sweden Austria Denmark Belgium Finland Luxembourg Bulgaria Iceland Belarus Russia Croatia Poland Spain � › Seven large net importers of data in EU/ ETFA region › In Europe most outbound traffic amounts to no more than ~ 10% of available link capacity › How does this compare with the Internet generally ?
Utilization › The backbones of the Internet are run at 10% to 15% of their capacity › Private line networks are utilized 3% to 5% . › low utilization of data networks is not a symptom of waste. › Low utilization rates lead to great opportunities for higher quality or less expensive service from aggregation of traffic. SOURCE: Andrew Odlyzko, University of Minnesota Data netw orks are lightly utilized, and w ill stay that w ay Review of Network Economics, 2 (no. 3), September 2003, pp. 210-237 http: / / www.dtc.umn.edu/ ~ odlyzko/ doc/ networks.html
Com pendium Survey on QoS › Does the NREN offer the sam e levels of QoS on the netw ork as those offered by GÉANT2 ? › IP Best Efforts › IP Less than Best Efforts › Premium YES 2 4 % NO 7 6 % 7% NREN hardware is not capable 21% NREN sees no demand for these services 4% Not physically possible unless all domains in path participate 4% Not economically viable 57% Prefer to over-engineer the network 12% Other reason n= 37 Slide 10
W hy low utilization is necessary › Low utilization comes from different patterns of use, lumpy capacity of transmission facilities, and the high growth rate of the industry Lightly loaded Saturated › Users value the ability to send data in high speed bursts, and that should guide us in the design and operation of networks › Also need to address end-to-end performance The last mile – application tuning… etc,
NREN Traffic to and from Com m ercial T3 T4 I nternet 2 0 0 7 100 90 80 70 60 50 40 30 20 10 0 Com m ercial I nternet Sites % of T3 ,T4 traffic to/ from
Traffic Sources and Destinations NREN sites External com m unity T1 + T2 Traffic T3 + T4 to/ from Other Global I nternet › Traffic to/ from global I nternet is legitim ate › NRENs may allow content providers to locate servers on their network to improve access to content › Aggregation of Global I nternet traffic and procurem ent of peering m akes econom ic sense.
Total I P traffic grow th on GÉANT 2 0 0 4 -2 0 0 8 Total IP traffic growth on G É ANT 2004 ‐ 2008 14000 12000 10000 Tbyte per month 8000 total IP 6000 E xpon. (total 4000 2000 0 F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ 04 04 04 04 04 04 05 05 05 05 05 05 06 06 06 06 06 06 07 07 07 07 07 07 08 08
Total I Pv6 traffic grow th on GÉANT 2 0 0 4 -2 0 0 8 IPv6 g rowth on G É ANT 2004 ‐ 2008 140 120 100 80 Tbyte per month IP v6 60 40 20 0 F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ Jun ‐ Aug ‐ Oct ‐ Dec ‐ F eb ‐ Apr ‐ 04 04 04 04 04 04 05 05 05 05 05 05 06 06 06 06 06 06 07 07 07 07 07 07 08 08
Total I P and I Pv6 traffic grow th on GÉANT 2 0 0 4 -2 0 0 8 Total IP & IPv6 g rowth on G É ANT 2004 ‐ 2008 13000 12000 11000 10000 9000 8000 Tbyte per month 7000 6000 5000 4000 3000 2000 1000 0 F eb ‐ 04 May ‐ 04 Aug ‐ 04 Nov ‐ 04 F eb ‐ 05 May ‐ 05 Aug ‐ 05 Nov ‐ 05 F eb ‐ 06 May ‐ 06 Aug ‐ 06 Nov ‐ 06 F eb ‐ 07 May ‐ 07 Aug ‐ 07 Nov ‐ 07 F eb ‐ 08 May ‐ 08
I Pv6 as a percentage of all I P traffic IPv6/(IPv4 + IPv6) percent 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% Mar-04 May-04 Jul-04 Sep-04 Nov-04 Jan-05 Mar-05 May-05 Jul-05 Sep-05 GÉANT: Percentage IPv6 traffic Nov-05 Jan-06 Mar-06 May-06 Jul-06 Sep-06 Nov-06 Jan-07 Mar-07 May-07 Jul-07 Sep-07 Nov-07 Jan-08 Mar-08 May-08
EU/ EFTA NREN Funding Sources Non-User/ Client Funding LONG TERM horizon User/ Client Funding SHORT TERM horizon
Elem ents of NREN Activity Production I nnovative NREN Com m odity Services Developm ent I NDI RECT DI RECT SPI LLOVER VALUE VALUE VALUE TOTAL NREN COSTS NREN Users/ Clients see VALUE PUBLI C VALUE User funding appropriate Central funding appropriate Acknow ledgem ents to: Donald Clark, REANNZ
Relating Reality to Econom ic Theory Sim plified 2 0 0 7 EU/ EFTA NREN Principal-Agency Theory Funding Sources Transaction-Costs Econom ics 1 Percentage User Charging 0 % 1 0 0 % Percentage User Charging optim al outcom es sub-optim al outcom es 1 0 0 % Central 1 0 0 % User Funding Funding Optim al ratio ? 1 ) Acknow ledgem ents to: Donald Clarke, REANNZ
Scenarios › Regulatory › Com m ercial / Com petition issues › User/ bill-payer funding issues › Lack of Political Support Slide 21
Regulatory › Cooperative relationship › Exam ple: FUNET › Converse › Exam ple: SURFnet › I ssues: › Requirem ents for data collection/ retention and providing taps for agencies › NRENs are not public netw orks › Closed user groups w ith lim ited scope › Need the Freedom to I nnovate successfully Slide 22
Com m ercial / Com petition I ssues › No serious incidence of problem s to date › NREN Position: › NRENs are not open public netw orks › Closed user groups w ith lim ited scope ( R&E) › Occupy a niche not served com m ercially › I nnovation for tom orrow s I nternet › Experts at integration of existing products into new and innovative pilot services › Cooperative w ith I ndustry for m utual benefit › Testbeds, equipm ent testing › Trickle dow n to com m ercial w orld and e-com m unity Slide 23
Recommend
More recommend