The Direction of Value Flow in Multi-service Connectionless Networks Bob Briscoe BT Research 7 Oct 1999
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value2 menu 4 why think about this? 4 the value of place 4 end-end pricing 4 common case value apportionment 4 end-end vs. iterative clearing context
pricingclearingscenariossummary 7 Oct 1999Direction of Value3 why think about this? 4 theory of the value of communications: –connection-oriented hegemony [ITU96, Zull97, Clark96, MacKie-Var92] 4 confusion is abroad in the world –junk advertising, bundling content+network 4 why assume multicast sender pays? 4 alternative proposal seems obvious –but wasn't when I started context
pricingclearingscenariossummary 7 Oct 1999Direction of Value4 context 4 multiservice, multicast IP (real-time apps) 4 some services usage-charged, whether packet, flow or even longer term granularity 4 electronic tariff announcement 4 buyer's market 4 designed for bundling –will cover in a future paper... context
pricingclearingscenariossummary 7 Oct 1999Direction of Value5 what business? 8 link (wholesale cost) 4 network 8 connection? (not supplied) applic’n present’n session transport network link physical network link physical end-system router network link physical applic’n present’n session transport network link physical context e-commerce
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value6 the value of place 4 xmt value /leg = (value of place/time) 4 but competition cost+ margin pricing 4 who can know the value of place? N: provider Y: end-users – apportionment 4 clearing grows comms market i i value 1 = f(i, place 1, time 1 ) value 2 = f(i, place 2, time 2 ) i context
pricingclearingscenariossummary 7 Oct 1999Direction of Value7 top level architecture! rudimentary interface no use alone customer provider eg. unicast to customer of same provider pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value8 end to end pricing price end-to-end pricing role service data flow customer A C ISPs B pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value9 end to end pricing end-to-end pricing role price service data flow pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value10 service service bundling data flow pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value11 end-to-end pricing role 4 not necessarily a business; a role taken by: –one ISP –the customer –a third party (+ bundling content?) 4 but…...what is the best default apportionment?... pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value12 default value apportionment 4 proposal: –'sender and receiver both charged' –slight discrepancy in prices possible –every leg of every flow is part charged twice –model self-similar for inter-domain boundaries pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value13 split-edge pricing - per direction 4 price in & out separately 4 each price between boundary and remote edge 4 price effects localised 4 contracts localised 4 extends edge-pricing [Shenker96] NaNa NbNb NcNc NcNc NdNd NdNd NeNe W bas W abr W abs W bar pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value14 split edge pricing-per class 4 each class of service has a market 4 class mismatch effects localised pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value15 split-edge pricing-per leg V QuQuQuQu QtQtQtQt QuQuQuQu QuQuQuQu QvQvQvQv QwQwQwQw QwQwQwQw NaNa NbNb NcNc NcNc NdNd NdNd NeNe W abr W bas W abs W bar 4 multicast & heterogeneous QoS pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value16 half-circuit charging 4 covers costs on average 4 separate rcv & send prices fit asymmetric costs pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value17 default value apportionment 4 choose default to minimise clearing 8 'only senders' or 'only receivers' charged 4halve transaction volume 8require industry-wide co-operation 8unstable - mavericks gain and increase costs of all 8multicast requires trusted receiver count 4 'sender and receiver both charged' 4majority of comms between consenting parties 4value, charge and cost congruent 4stable against maverick policies 8doesn't match blame for transmission... pricing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value18 blame, liability and control 4 but…...sender always to blame for a transmission 4 request-reply is irrelevant to network layer 4 other confusions clarified: value of moving info transitory value of moving GBs is often for a few bits cost of xmt << cost of targeting what to xmt clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value19 dilemma & solution 4 'sender and receiver both charged' vs. 'sender to blame' 4 receiver customarily pays...but... sender liable 4 cf. postage stamp sender customarily pays, but receiver liable clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value20 clearing across edge providers third party clearing service money clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value21 clearing across edge providers third party clearing service money clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value22 clearing across edge providers iterative service money 54 clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value23 end-to-end model: justification 8 avoids Int'l Accounting Rate System flaws: 8muddling wholesale with clearing 8per-flow metering in backbone 8trusted end counting for multicast 8apportionment needs authentication therefore unaggregateable 8ac d+1 prices in system 6e26 clearing c 20,000 d 5 a = apportionment schemes per ISP 10 a = apportionment schemes per ISP 10
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value24 end-to-end model: justification II 4 localised contract jurisdiction clear what to meter decided locally 4 selective openness of clearing process 8 ac 2 prices problem still exists 4e9 clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value25 customer bundling model 8 customer bundles together providers on end-to-end path 8 per-flow pricing 8high latency 8 contractual relations with arbitrary remote parties clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value26 exception peering 4 exceptional failures –loss of a packet subject to reservation –excess delay of a low latency packet 4 no customer revenue 4 no recriminations between ISPs –no “who lost packet?” –no “who delayed packet most?” 4 assume peers sufficiently dimensioned clearing
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value27 scenarios 4 see paper for... 4 finding an end-to-end price 4 sender liable but local payment customary 4 inter-domain multicast with heterogeneous quality of service (QoS) 4 phone to Internet gateway (PIG) 4...ie. per session accounting scenarios
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value28 PIG - phone to Internet gateways service money Internet end-end PSTN iterative clearing PIG scenarios
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value29 alternative now seems obvious... 4 per-session accounting bill=anachronism 4 split-edge pricing scales infinitely 4 'sender and receiver both charged' stable –customary - but only sender liable to pay 4 end-to-end pricing and clearing efficient 4 end-to-end intermediary could take 0-50% of retail ISP and telco markets –...and their customer relationships summary
contextpricingclearingscenariossummary 7 Oct 1999Direction of Value30 further information 4 Differential charging (diffchar) project Bob Briscoe more info