Service Ecosystems Management for Collaborative Process Improvement

[Print-friendly version]

ARC Linkage Grant: LP0669244

Industry Partners:

• SAP Australia (SAP Research, Brisbane)
• Queensland Government, Department of Public Works

Period: 2006 - 2009

Budget: AU$ 499,000 ($259,000 ARC + $120,000 SAP + 120,000 Queensland Government)

Investigators:

• Professor Michael Rosemann
• A/Prof Marlon Dumas
• Prof Amanda Spink
• Prof Peter Bruza
• Dr Alistair Barros (SAP)
• Dr Paul York (Queensland Government)
• Lawrence Lim (PhD student)
• Christian Prokopp (PhD student)
• Joern Franke (German Masters student)
• Christoph Riedl (German Masters student)

Category: 2801 - INFORMATION SYSTEMS

Summary:

Streamlining cross-organisational processes based on service-oriented architectures is perceived as a natural way of increasing organisational performance. However, an overarching framework for nurturing the orderly development and deployment of service-oriented systems in large and highly compartmentalised organisations is missing. This project seeks to provide methods for consistent definition of services, design of user-centred service repositories, incentive and accountability structures for service governance, and efficient service discovery mechanisms in service-oriented systems. Among other outcomes, the project will lead to an overarching framework for enabling the formation of so-called service ecosystems, that is, service-oriented systems that grow up from independent initiatives, yet interact in seamless ways to support complex business processes. Though the focus will be on government processes, the project will generalise beyond this scope for a wider uptake of service ecosystems in other types of organisations.

The project is divided in the following five work packages:

Service Analysis and Design (Marlon Dumas)

Service Analysis and Design (SAD) refers to the stages in the lifecycle of a service-oriented architecture where the purpose, scope and interfaces of individual services are identified, and an priori relationships between services are documented. It is also in these stages that the relationshops between service models and other enterprise architecture artifacts, such as organisational charts, document flow charts and business process models are specified. Various methodologies for SAD have been proposed. However, there are no theoretical or empirical studies to back their applicability, their effectiveness and their relative trade-offs. Also, it is unclear that these methodologies address the requirements of 'federated organisations' consisting of highly independent units such as Queensland Government (QG). In such environments, it is important that the methodology can be applied by independent teams in a way that leads to consistent and uniform results. Indeed, if different agencies fail to define their services consistently and uniformly, it becomes difficult to reuse and interconnect these services. For example, it is undesirable that one agency views a service as a permitting service while another agency views it as a reservation or a sales service. Finally, it is unclear that existing SAD methodologies are compatible with the vision of a service ecosystem where services are delivered through multiple channels and intermediaries, with varying pricing models, payment methods, authentication and authorisation mechanisms, etc.

Although SAD encompasses a wide range of activities, including analysis of non-functional requirements (e.g. security and quality of service requirements, service level agreements, outsourcing decisions, etc.), this work package will focus on two aspects: (i) identification and decomposition of services; (ii) identification and documentation of service interactions; and (iii) alignment of service interaction models with business process models. The research questions are deliverables outlined below must be understood in light of this chosen aspects.

This package will address the following research questions:

1. What are the underlying assumptions, the relative advantages and disadvantages of existing SAD methodologies?
2. Are these methodologies applicable in federated environments and if not, can they be extended, adapted and/or combined to address the needs of federated environments?
3. Are existing SAD methodologies compatible with the notion of service ecosystem and if not, how can they be extended, adapted and/or combined to address the needs of service ecosystems?

The deliverables of this package are

• D1 (month 3): A document outlining the state-of-the-art of existing SAD methodologies and a set of hypothesis regarding their relative tradeoffs.
• D2 (month 9): A document and set of models obtained from one or two explorative case studies where some of the SAD methodologies studied earlier are applied to at least one agency or project with QG
• D3 (month 12): A document outlining conclusions regarding the relative tradeoffs between existing SAD methodologies and their applicability in federated environments, and the possibility of extending, adapting and/or combining methodologies to address any shortcomings identified in deliverable D3.
• D4 (month 15): A document detailing an extended, adapted or unified methodology for SAD addressing any shortcomings identified in deliverable D3.
• D5 (month 21): A document and set of models obtained from one or two validating case studies in which the methodology of deliverable D4 is applied to at least one agency or project within QG.
• D6 (month 24): A document detailing a revised version of the methodology of deliverable D4 taking into account lessons learned from the case studies documented in deliverable D5.
• D7 (month 33): A document outlining difficulties in applying the methodology of deliverable D6 in the context of large and dynamic service ecosystems, as well as ways of addressing these difficulties.
• D8 (month 36): A document outlining an extended, adapted or unified methodology for SAD addressing any difficulties identified in deliverable D7.

Service Discovery (Peter Bruza and Alistair Barros)

At first sight, web service discovery would seem to be addressed by state-of-the-art information retrieval technology: The textual descriptions of services are treated as documents, the agents' agenda is described by a query, and the IR system ranks service descriptions on decreasing relevance to the query. Those highly ranked description will presumably link to services which are relevant to the agenda the agent wishes to close. One can describe this as the "Service-Google" approach.

In previous research we hypothesized the Service-Google approach starts to degrade when the associated agenda becomes more sophisticated. Crudely put, for more complex agendas with varying facets, the user is unlikely to know what they need to know in order to issue appropriate queries, or for some facets, they may not be able to issue any queries at all. By way of illustration, imagine your agenda is to set up a coffee shop in Brisbane, there are likely to be issues in relation to this agenda which of which you are totally ignorant. Consequently, our view of web service discovery is not only about the retrieval of relevant web-based services but also helping the user discover what it is (s)he needs to know in order to close the agenda at hand. To borrow, and slightly abuse "introspection", a term from epistemic logic, technology should aid the user to become both "positively" and "negatively" introspective about their agenda. For this reason, and others, our earlier research has focussed on technology which operationally replicates human pragmatic inference, like abduction. Such technology attempts to produce suggestions about issues relevant to the agenda of which the user may not be aware. In this way the user is "triggered" into introspection and thus learns about the agenda and associated problem space (positive introspection) or what they don't know (negative introspection) as they go along. In this sense, the technology we propose is essentially an exploratory search system. In this project, we aim to further develop the exploratory search approach to web service discovery but in relation to specific contexts and situations relevant to the Queensland Government.

Some motivating research questions:

1. Does the Service-Google approach suffice for the Queensland Government? If not, is it possible to characterize when it does suffice, and when it does not?
2. Have the British empiricists got it right in the context of web service discovery? Is it sufficient to learn by association in relation to the agenda at hand? If so, from where should the technology prime the associations, e.g., cognitive information structures (semantic space models), normative information structures (legislation, taxonomies, ontologies), social information structures (folksonomies)?
3. How can aspects of context be brought in to condition the interaction between user and service space in order to facilitate issues such as relevance, learning, abductive triggers, and ultimately effective discovery.
4. Does service discovery include the problem of discovering what services ought to exist, but don't?
5. Service "crawling" - what service are "out there"?
6. Under the assumption that web service discovery involves high levels of contextual dependency, it is a complex system. This is in line with the ecosystem philosophy, as ecosystems are complex systems. This being so, a reductionist approach won't work. Traction may be gained by viewing services, not as objects, but as interactions whereby (some) relevant properties only manifest at the point of interaction. This view not only poses intriguing challenges for service discovery, but the whole question of adequately modelling services and service ecosystems.

Service Value Management (Amanda Spink)

Service value management (SVM) refers to the stage in the life cycle of a federated service-oriented architecture where the service repository is evaluated. Various approached to service value management have been proposed. However, few empirical studies have proposed comprehensive approaches to service valuve management evaluation and usability. The Enterprise Systems Success (ESS) Measurement Model (Gable, Sedera & Chan, 2003; Sedera & Gable, 2004) provided a validated measurement model and instrument for assessing enterprise systems success.

This package will address the following research questions:

1. What are the relative advantages and disadvantages of existing approaches to service value management (SVM)?
2. Is the Enterprise Systems Success (ESS) Measurement Model applicable for the federated environment?
3. Is the EES Measurement Model compatible with a service ecosystem, and if not, can it be extended, adapted and/or combined to address the need to evaluate the service ecosystems?

The deliverables of this package are:

• D1 (month 6): A document outlining the state-of-the-art of existing approaches to service value management evaluation.
• D2 (month 12): A document and set of models obtained from an exploratory case study of where some SVM evaluation approaches studied earlier are applied to one agency or project within QG.
• D3 (month 18): A document detailing an extended, adapted or unified approach to service value management.
• D4 (month 24): A document and models obtained from a validating case study applied to at least one agency or project within QG.
• D5 (month 33): A document detailing a revised version of the methodology in the context of large and dynamic service ecosystems.
• D6 (month 36): A document outlining an extended, adapted or unified methodology for SVM addressing any difficulties identified in D5.

Service Governance (Michael Rosemann)

Service governance refers to the decision making processes and the corresponding roles and responsbilities which complement the lifecycle of a service-oriented architecture. As such it provides a shell of accountability for service management. Governance in the context of service ecosystems can be discussed on different levels, covering service lifecycle governance, service composition governance and service ecosystem governance. Service governance has also to be integrated with other forms of governance such as corporate governance or IT governance. Furthermore, governance is impacted by demographics such as public/private sector or centralized/federated or decentralized structures which will impact relevant policies and procedures.

Various methodologies for governance have been proposed, especially for corporate governance and IT governance, and these will impact the initial design of service governance models. Typically governance is driven by the desire to increase conformance and performance. However, there are no dedicated studies on the specific requirements of and approaches for the emerging area of service governance, let alone service ecosystems governance.

The research on service governance in this project will in analogy to similar work on IT governance (Peter Weill) explore the main decisions and relevant roles (e.g. service architect) in the area of service ecosystems management. This will govern all stages of the service lifecycle incl. service composition. A focus will be on the effective accountability structures and incentives. Well-defined accountability will assign ownership to services and service ecosystems which is a pre-requisite for its integration into the overall (IT) management. Incentive schemas will encourage the continuous population, use and maintenance of the service ecosystem. The research on incentives has to take restrictions within the public sector into account, and will borrow concepts from incentive schemas has they have been designed within knowledge management.

This package will address the following research questions:

1. What are the underlying assumptions, the relative advantages and disadvantages of existing service governance methodologies, especially accountability structures and incentive schema?
2. How can the set of roles and processes relevant for service governance be identified?
3. Are these methodologies applicable in federated environments and if not, can they be extended, adapted and/or combined to address the needs of federated environments?
4. What specific constraints (if any) does the domain of an Australian State Government impose on service governance, especially accountability structures and incentive schema?
5. Are existing IT/service governance methodologies compatible with the notion of service ecosystems and if not, how can they be extended, adapted and/or combined to address the needs of service ecosystems?

The deliverables of this package are:

• D1 (month 3): A document outlining the state-of-the-art of existing service governance methodologies and a set of hypotheses regarding their relative tradeoffs. This document will also define elements out-of-scope for this project but with potential impact on service governance.
• D2 (month 9): A document and set of models obtained from one or two explorative case studies where some of the service governance methodologies studied earlier are applied to at least one agency or project with Queensland Government. This deliverable will also discuss the identified roles and processes related to Service Governance.
• D3 (month 12): A document outlining conclusions regarding the relative tradeoffs between existing service governance methodologies and their applicability in federated environments, and the possbility of extending, adapting and/or combining methodologies to address any shortcomings identified in deliverable D3.
• D4 (month 15): A document detailing an extended, adapted or unified methodology for service governance addressing any shortcomings identified in deliverable D3.
• D5 (month 21): A document and set of models obtained from one or two validating case studies in which the methodology of deliverable D4 is applied to at least one agency or project within QG.
• D6 (month 24): A document detailing a revised version of the methodology of deliverable D4 taking into account lessons learned from the case studies documented in deliverable D5.
• D7 (month 33): A document outlining difficulties in applying the methodology of deliverable D6 in the context of large and dynamic service ecosystems, as well as ways of addressing these difficulties.
• D8 (month 36): A document outlining an extended, adapted or unified methodology for service governance addressing any difficulties identified in deliverable D7

Service Delivery and Contextualisation (Alistair Barros)

A key value-proposition of service ecosystems is enabling services from wide-ranging sources cutting private and public sectors to be traded outside their traditional custodial and governance boundaries into new and unforeseen markets. Beyond remote access and composition of service endpoints, as advanced through service-oriented architectures and core Web services standards, the wider setting of service ecosystems implies that aspects of service provisioning can be outsourced to third parties for new business opportunities.

A prominent aspect of outsourced service provisioning involves service delivery. In service industries, new channels for service delivery are established through service brokers wherein services are published so that they may be discovered, ordered, authorized for access, paid for and utilized by customers. In the public sector, information brokers, integrated service desks and trusted agencies (e.g. Australia Post) provide such "front-desk" support. In the private sector, banking and insurance industries outsource sales and ordering of their services through individuals and business partners.

Meanwhile on the Web, brokerage has gained prominence through digital content (e.g. iTunes) and software-as-a-service (e.g. Salesforce and StrikeIron) marketplaces. With the rapid success of these ventures, the question has turned to how current technical solutions for brokerage scale for different sorts of services. In iTunes, Salesforce and the like, the form of service delivery has similarities with conventional goods procurements given that the consumable items are much like goods: music, video and software widgets.

In considering business services of especially government, financial, health, and legal domains, current provisions for brokerage support present limitations. Services in these sectors are not amenable for pay/download since they are typically implemented through legacy applications hosted by providers. Thus, consumer access to services entails interactions with their interfaces and backend implementations. Indeed, several and perhaps a long-running set of interactions might be involved, and more complex services may entail several stakeholders and implementations through long-running business processes.

For third-party brokers to develop new markets for business services and to deliver these, special service interfaces allowing new contexts of service delivery are required without violations to service access constraints put in place by service providers. As an example, brokers may introduce advertising content and other discount mechanisms into services, reducing or neutralizing their costs for consumers while preserving revenue flow back to providers. Thus, a new layer of service interfacing is required which enables services to be placed into new contexts of service delivery. Such interfacing should be carefully integrated with backend service implementations hosted through governing service providers.

Key research questions:

1. What are key aspects of service delivery that can be re-contextualized and how can these be described through new layers of service interfaces? In other words, how can new aspects service delivery, such as advertising add-on, payments, authorization and usage metering, be introduced into new service interfaces?
2. In allowing aspects of service delivery to be configured into new service interfaces, what are the reengineering implications for backend service implementations? In general, how should such a service be structured and integrated with service implementations?
3. How can new service interfaces for service delivery be efficiently captured and analysed for compatibility, both structurally and behaviourally, with respect to their backend implementations? How can incompatibilities be overcome through adaptations? To what extent can such adaptations be supported?

Related Publications

Alistair Barros, Marlon Dumas and Peter Bruza
The Move to Web Service Ecosystems. (PDF, 272Kb)
In BPTrends, November 2005