Presentation on theme: "June 27, 2007 FIND Meeting, 2007 1 From Packet-Switching to Contract- Switching Aparna Gupta Shivkumar Kalyanaraman Rensselaer Polytechnic Institute Troy,"— Presentation transcript:
June 27, 2007 FIND Meeting, 2007 1 From Packet-Switching to Contract- Switching Aparna Gupta Shivkumar Kalyanaraman Rensselaer Polytechnic Institute Troy, NY Murat Yuksel University of Nevada – Reno Reno, NV
June 27, 2007 FIND Meeting, 2007 2 Implied Challenges Motivation Current problems: Users cannot express value choices at sufficient granularity – only at access level Providers do not have economic knobs to manage risks involved in investing innovative QoS technologies and business relationships with other providers flexibility in time: forward/option pricing flexibility in space: user-defined inter- domain routes capability to provide e2e higher quality services money-back guarantees, risk/cost sharing
June 27, 2007 FIND Meeting, 2007 3 Contract-switching: A paradigm shift… Circuit-switching Packet-switching Contract-switching ISP A ISP C ISP B e2e circuits ISP A ISP C ISP B routable datagrams ISP A ISP C ISP B contracts overlaid on routable datagrams
June 27, 2007 FIND Meeting, 2007 4 Basic Building Block: Intra-domain dynamic contracts An ISP is abstracted as a set of contract links Contract link: an advertisable contract between peering/edge points i and j of an ISP with flexibility of advertising different prices for edge-to- edge intra-domain paths Contract components Performance component Time component Financial component
June 27, 2007 FIND Meeting, 2007 5 A Contract-Switched Network Core Contracts: a practical way to manage value flows Technologies to support QoS Economic considerations for service definition and delivery Scalability, Efficiency and Fairness Contract timescales Cost recovery Pricing the risk in QoS guarantees Single-domain and end-to-end contracts
June 27, 2007 FIND Meeting, 2007 6 Pricing End-to-end QoS Contracts End-to-end contract characterized by source-destination (s-d) pair other specifications, eg. QoS specs, contract duration Two-component pricing model (P e = P bw + V*) P bw component for cost recovery (single domain and e-2-e) V* component for risk management of QoS assurance provides appropriate scaling between P bw and V* Balance between customer demand for vanilla bandwidth and additional QoS assurance Determined by cross sensitivity between demand for vanilla bandwidth and additional QoS guarantees Develop to handle complexity and offer efficiency - improve profitability, risk sharing, customer welfare, and utilization
June 27, 2007 FIND Meeting, 2007 7 Pricing Bandwidth for Cost Recovery Nonlinear pricing model to recover providers cost Bandwidth purchase cost from constituent ISPs Fixed cost to setup and maintain transit nodes Price schedule responds to customer demand Categorization based pricing for complexity management Distance from s to d: hop counts h Speed of traffic from s to d: bottlenecks b Bandwidth pricing problem:
June 27, 2007 FIND Meeting, 2007 8 Pricing of Risk in End-to-end QoS Guarantee Single-domain contracts stitched to create end-to- end QoS assured contracts Risks in end-to-end QoS assurance from Constituent contracts Stitch nodes Risk management using pricing Contract with N ISPs Intra-domain contracts specified with End-to-end contract Definition of end-to-end contract (QoS assurance) Pricing strategies
June 27, 2007 FIND Meeting, 2007 9 Model for Pricing Risk in End-to-end QoS Price specified by contract: sd pair QoS (Loss) guarantee Temporal characteristics, etc determined by lowest price over all likely concatenations to deliver between sd pair
June 27, 2007 FIND Meeting, 2007 10 Putting it together – Contract switching, Routing, Financial Engineering End-to-end QoS services Contract Routing Pricing Risk management tools Spot contracts Forward contracts Options on Forward Flexibility to innovate services
June 27, 2007 FIND Meeting, 2007 12 Definition of End-to-end Loss Guarantee Type of contract The per minute loss rate of the customers data over contract duration T starting from t 0 does not exceed Constituents of end-to-end loss Definition of end-to-end loss guarantee
June 27, 2007 FIND Meeting, 2007 13 Pricing of Risk in Loss Guaranteed Intra- domain Services Sample Contract: The per minute maximum loss rates are less than 0.5% (S i u ) over the contract duration of 1 hour. Per Minute Loss Rate l t : Provision of loss based QoS guaranteed services is risky. Due to the uncertainties caused by the competing traffic. Outcome of loss process in favor of or against the provider.
June 27, 2007 FIND Meeting, 2007 14 Pricing of Risk in Loss Guaranteed Intra- domain Services Payoff defined as where is the upper barrier (providers promised loss rate guarantee), and is the indicator function defined as Price for the risk: where -- total number of minutes of the contract duration, -- the risk neutral measure from providers SPD.
June 27, 2007 FIND Meeting, 2007 15 Pricing of Risk Using State Price Density Price of Risk needs to be assigned for unhedgeable risk. State Price Density (SPD) (3) SPD describes a representative providers preferences for the future outcomes of the loss process. Assumptions of the providers preference: The provider would expect losses to be rare events. The provider would not get rewarded for large losses. Two alternative forms of SPD functions: A monotonously decreasing SPD. A SPD peaking at a positive loss rate.
June 27, 2007 FIND Meeting, 2007 16 Constructing a State-Price Density p1 p2 p3 T=0 T=1 0 Mb 1 Mb 100 Mb 5c 1c 0c 5/6 1/6 0/6 Ten such time steps with (8, 1, 1) realization of each outcome imply a value of 8*5/6 + 1*1/6 + 1*0/6 = 41/6.
June 27, 2007 FIND Meeting, 2007 17 Sample Choice of State-Price Densities Study price evolutions with different SPDs Network settings Capacity Customers traffic I t the Aggregate A t Sample SPDs SPD 1: Exp(0.02) SPD 2: Beta(1.5, 100.5) SPD 3: Beta(1.5, 167.2) SPD 4: Beta(1.05, 100.95)
June 27, 2007 FIND Meeting, 2007 18 Price Variations with Different SPDs A decreasing SPD (SPD 1) produces performance based prices. A SPD that does not reward zero losses produces congestion sensitive prices. Among the beta SPDs, the SPD that rewards higher for smaller losses is more favorable to the provider.