Share
Embed
UCC 2013 Keynote Christof Weinhardt: Networked Services - The Meeting Point of Computer Science and Economics
Christof Weinhardt, Professor at the Karlsruhe Service Research Institute (KRSI) of the Karlsruhe Institute of Technology (KIT) talking at the 6th IEEE/ACM International Conference on Utility and Cloud Computing in Dresden, Germany
ucc2013
December 10, 2013
Transcript
41 slides
-
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Karlsruhe
Service Research Institute (KSRI) Prof. Dr. Christof Weinhardt www.kit.edu NETWORKED SERVICES The Meeting Point of Computer Science and Economics 6th IEEE/ACM International Conference on Utility and Cloud Computing UCC Dresden, December 10th, 2013 -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Overview
1 Service Value Networks: Terms and Foundations 2 Market Engineering for Service Value Networks 3 Mechanism Design for Networked Services 4 Conclusions and Outlook -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu From
Service Mashups to Service Networks Service Oriented Architectures Service Mashups Service Value Networks From monolithic software applications to Service Oriented Architectures (SOA) Modularity is the key to face rising demands for sophisticated, customized services Customers purchase services on demand. Services can be dynamically composed from the offers of different modular providers. -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Mashups Data/API-calls from different sources Service Mashup Goals Information § Combination § Visualization § Aggregation -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Mashups: Critique Data/API-calls from different sources Service Mashup Static Composition: Exchange of services not easily implemented Service Evolution: Mashup not easily adaptable to new technologies, more complex services, etc. -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Value Networks Fully customized solutions from one provider or intermediary IT as an enabler Providers focus on core competencies Dynamic, automated composition Service Value Networks -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu I
want to support a business process which comprises of Payment and Billing. Licence-based software? Software as a Service? Service Park, Service Value Network (SVN) Orchestration platform with common, specified standard Adaptability, Specificity? Service Value Networks Specificity through flexible (re-)combination of service modules Service Value Networks are smart business networks that provide business value by performing automated on- demand composition of complex services from a steady, but open pool of complementary as well as substitutive standardized service modules through a universally accessible network orchestration platform. -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Park1 § An established (web) service community with a common runtime § Park operators set: § Rules of Engagement, Participation etc. § Business Objects § Pricing and Allocation Mechanisms § Entrance Fees and Requirements § Parks are feasible due to their constrained and centralised approach Service Park, Service Value Network (SVN) Orchestration platform with common, specified standard 1: [Petrie and Bussler 2008] -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Evolving
a Service Value Network What are suitable incentives for attracting and retaining providers in a Service Value Network? What are the key economic institutions in engineering a sustainable platform for the dynamic composition of complex services? -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Requirements
for Network Formation Network Development Network Formation and Growth Network Administration and Profitability Phase Goals Business Model § Increase network size and service diversity § Attract new providers and customers Ensure market sustainability § Profitable platform § Participation incentives § Reward participation in a fair way § Accept temporary non- budget-balance § Ensure budget-balance § Limit effects of missing incentive compatibility -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Dr.
Simon Caton Market Engineering -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
providers § Control the services § Private information § Preferences § QoS attributes § opportunistic Service requester § Requires complex service § Willingness to pay § Preferences on Quality-of-Service (QoS) attributes Platform operator § Utility maximization over all participants § Incomplete information § Preferences § QoS attributes Engineering Problem(s) à Definition of a social choice function [Hurwicz 1975, Green & Laffont 1979, Myerson & Satterthwaite 1983] Mechanism Design Challenges Information asymmetry and opportunistic behavior M -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Value Networks Mapping of specific customer requirements I want to support a business process which comprises of Payment and Billing. Billing Payment Price, Quality -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Value Networks Graph-based formalization Billing Payment -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Value Networks Graph-based formalization -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Value Networks Graph-based formalization Scoring function (“fit”) Complex service -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu à
Complex Service Auction -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Payment
Billing Complex Service Auction for dynamic allocation (joint work with Benjamin Blau) 20 I am offering my service 4 with a customer satisfaction of 1.0 for a price of… • … 17 as successor of sevice 1, and … • … 20 as successor of service 3. I need a complex service consisting of Payment and Billing, and I am willing to pay up to 100 for it. -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Complex
Service Auction for dynamic allocation UT = 100 (min{0.6,0.7}) – (13 + 16) = 31 UM = 100 (min{0.6,1.0}) – (13 + 17) = 30 UB = 100 (min{0.5,1.0}) – (10 + 20) = 20 Provider of Service 1 gets: Provider of Service 2 gets: Customer utility: Willingness to Pay Allocation Function Transfer Function Legend Aggregation Function Throughput „Critical value“ Truth telling with respect to prices is weeky dominant strategy -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Complex
Service Auction Compensation and Service Level Enforcement § Service providers might have an incentive to lie about their services’ configuration § Idea is to distinguish between a: • Declaration phase Ex-ante announcements between requester and providers • Execution phase Ex-post verification of announcements during run-time (e.g. WS monitoring) “The payments need only to be given after the execution. Intuitively we view the execution part as allowing the mechanism to verify in some sense the agents’ declarations, and “punish” them for lying.” [N. Nisan et al. 2001] Compensation term: Maximum utility based on actual realized attribute values of service provider s’s services: Maximum utility based on announced attribute values: -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu The
Complex Service Auction (CSA) Ensuring SLA Adherence Service Level Enforcement Extension for CSA that enhances the transfer function by a compensation function, which • punishes service providers for untruthful announcements of QoS a>ributes and • compensates service requesters for the utility loss they incur due to resulting non-‐‑ or mal-‐‑performance Compensation Function * * * *, ervices owned by n : ( ) ( ) if , allocated : 0 otherwise k ver k ij k j ij S j t p v e t -‐ Ï Ô = + -‐ -‐ -‐ Ô Ô Ô =Ì Ô Ô Ô Ô Ó Â U U U U Truth-‐ telling with respect to all dimensions of the bid is a weakly dominant strategy Overall utility gap that results from the utility difference based on the announced a>ribute values and the verified ones measured after execution. { -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Complex
Service Auction (CSA) Extension Abstract model & Mechanism Design Blau et al (2009) Allocation Transfer Consumer surplus Provider surplus Critical value SLA verification Multidimensional incentive compatibility sacrifices budget balance. Hence, system requires constant subsidization! Sustainability? Budget Balance and Sustainability -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Achieving
budget balance – The Interoperability Transfer Function (ITF) 0,4 0,6 0,8 1 1,2 -‐‑25% 0% 25% 50% 75% 100% 125% 150% Mean provider utility relative to truth-‐‑telling utility Bid deviation relative to truth-‐‑telling ITF_12_4 ITF_16_4 ITF_32_4 Relative provider utility that deviates from the true valuation dependent on the degree of competition. ITF_{#Provider}_{#Cluster} Approximate incentive compatibility while retaining budget balance! • Computation of allocation and „critical values“ of service providers • Computation of approximate values that minimize the weighted sum of all distances to the actual „critical values“ in a budget-balanced manner (Parkes et al. 2001) Approach to Retain Sustainability [Parks et al.2001] -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Co-opetition
Mechanism for Networked Services -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Network
Formation NETWORK EFFECTS • Increasing returns with positive feedback loops • Co-opetition • Ecology thought: Participants ultimately share the fate of the network • è Two-sided Market MODULARIZATION/ SPECIALIZATION • Participants focus on core competencies • Leveraging the assets of partners as key to success in today’s dynamic environment • Providers should be flexibly interchangeable PERMEABILITY • Low entry barriers through loose coupling • The network is open to new entrants CUSTOMER-CENTRICITY • Offered through intermediaries that bundle modular service/product offerings into a complex, holistic package Participants Value Customers Value -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Service
Value Networks Challenges in networked economies Challenges • Two-sided markets: Critical mass • Co-opetition: Cooperative and competitive element in SVNs Co-opetition Mechanism Platform operator Design of a suitable coordination mechanism for SVNs in their launch phase Billing Payment -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu The
Co-Opetition Mechanism Allocation function Allocation function argmax argmax , : ( ) : ( ( ) ) l l l l ij l F F ij F F F F i j e E W o p Œ Œ Œ = = ◊ -‐  U Q A a Willingness to pay w.r.t. complexe service Fl Aggregated price bids of services in the complex service Fl Transfer function Compensation of allocation and the generation of a potential added value / contribution for the SVN -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu The
Co-Opetition Mechanism Contribution to SVN 2 15 F = U 1 20 F = U 3 25 F = U Contribution • If v4 fails, two other complex services are ready, which create positive utility for the system • Compensation of readiness to provide a service • Power Ratio quantifies the individual contribution of services to the SVN Internal cooperations and value function Network view Internal cooperations of a SVN Value function -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu The
Co-Opetition Mechanism Allocation and transfer function Approach: Incentivize via monetizing the created added value (Power Ratio). Transfer function (PRTF) tj for all services vj in V Willingness to pay w.r.t. complexe service Fl Allocation function Aggregated price bids of services in the complex service Fl Allocation-based and contribution- based component of transfer function else -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu The
Co-Opetition Mechanism Allocation and transfer function Transfer function (PRTF) tj for all services vj in V For all cooperations Sm , vj is a part of… … weighted by the probability of Sm to form,… … the power ratio calculates the added value (=marginal contribution) vj is responsible for . [Shapley 1953, Myerson 1977, Jackson and Wolinsky 1996, Jackson 2005] Approach: Incentivize via monetizing the created added value (Power Ratio). Willingness to pay w.r.t. complexe service Fl Allocation function Aggregated price bids of services in the complex service Fl Allocation-based and contribution- based component of transfer function else -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu (R1)
Budget Balance (R2) Individual Rationality (R3) Incentive Compatibility (R4) Allocative Efficiency (R5) Network Growth (R6) Readiness (R7) Fairness (R8) Interconnectedness (R9) Service Sequenz (R10) QoS AXributes (R11) SLA Enforcement (R12) Tractability 16.12.13 Classic mechanism design Network design Applicability in SVNs = fulfilled = parlty fulfilled = not fulfilled/ not applicable The Co-‐‑Opetition Mechanism Properties -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Summary:
Engineering Service Markets Automated service composition via Web APIs Social Business Network to augment SVN formation Auction mechanisms determine service collaborations Stimulating growth through co-opetition Service Value Networks -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Evolving
a Service Value Network Attract and retain providers in a Service Value Network? Economic institutions to create a sustainable platform? Networked Mechanism Design Complex Service Auctions -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Evolving
a Service Value Network Monitorable, fault tolerant and adaptive SVNs? Integrating social business relationships and trusted collaboration? ? -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Dr.
Simon Caton 238 Million Users from 200 Countries 3 Million Business Pages showcasing 1.2 million products and services 1 billion endorsements -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Can
someone complement my service? Me Me Social Cloud: Social Network-Based Collaboration and Exchange (Young Investigator Group – Simon Caton) Objective: leverage social networks for the collaborative exchange of resources, services and capabilities Approach: cross section of Computer Science (platform architectures and analysis technologies), Economics (market engineering and incentives) and Sociology (understanding online communities and their intricacies) -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Provisioning
and consumption of services Matching in Social Network-based Environments Economic and social matching mechanisms to allocate services -
Karlsruher Institut für Technologie Karlsruhe Service Research Institute www.ksri.kit.edu Christof
Weinhardt [email protected] Thank you for your attention!