Engineering complex service-oriented systems presents grand challenges due to their great complexity, volatility, and uncertainty in their rapidly evolving technology and social contexts. It demands an effective engineering approach in order to satisfy the needs of service economics. This paper proposes an approach called Organic Aggregation Service Engineering Framework (OASEF). Experience from proof-ofconcepts case studies shows that it provides a practical means to develop service-oriented systems. It enables and promotes a focus on higher-level intellectual engineering e orts, and provides a mechanism to capture and reuse engineering capacities in a model-driven environment.
Deriving and managing complex Service-Oriented Systems (SOS) presents great
challenges due to their nature and characteristics in a dynamically complex
environments [
Secondly, SOS often present a high degree of uncertainty. Autonomous
services are often provided and consumed by di erent agents for di erent types
of purposes, without necessarily knowing each other in advance [
These special characteristics and challenges can hardly be resolved by merely
adopting traditional software processes and engineering methodologies, or solely
relying on more advanced implementation technologies [
Therefore, our vision of an e ective service engineering approach should enable and support e cient and exible formation and exploitation of both higherorder intellectual resources and lower-order implementation processes. That is, both types of engineering resources should be explicitly identi ed, developed, captured, and used, as major engineering means, to produce and manage systems, in a exible, systematic, and automatic means, at least partially if processing in its entirety is not possible. Speci cally, our objective is to achieve an e ective service engineering approach that: i) promotes a focus on high-level intellectual activities within the world of human mind, such as exploring and understanding problematic situations, and identifying and capturing higher order engineering purposes and intentions; ii) links the captured higher-order engineering resources with other lower-order engineering activities by using the former to guide and shape the latter systematically and sensibly, within a coherent overall process; and iii) provides a means to capture, aggregate, and reuse these important engineering resources, including both higher-order and lower-order capacities, to facilitate exible and rapid aggregations of processes and systems.
This paper presents a new approach to service engineering based on such a vision, which is called Organic Aggregation Service Engineering Framework (OASEF). It is based on a philosophy that takes a synthetic approach to growing and managing processes and systems via sensible and responsive aggregations of resources and capacities. This work mainly has the following contribution: rstly, it emphasises exploration and exploitation of higher-order engineering e orts, and links them with lower-level technological resources within an overall process model; secondly, it promotes a concept of organic aggregation in engineering that captures engineering capacities as reusable resources, and, more importantly, uses them in a sensible, agile, and controlled fashion.
The remainder of the paper is organised as follows: section 2 presents the proposed OASEF framework in details, including a general engineering framework it conforms to, its conceptual model, and modelling methodology; section 3 brie y illustrates its important concepts and methods using results from two proof-ofconcept case studies; section 4 analyses its features and limitations based on observation and experience during the case studies, and compares it with some related work; section 5 concludes this paper with a summary. 2
It is important for a discipline of service engineering to have a well-founded engineering framework that coherently organises engineering activities and resources according to the nature and characteristics of services and systems under consideration. The aim of this work is to provide and assess such a framework. 2.1
The design of OASEF started from generalisation of other engineering
disciplines. Figure 1 depicts a general conceptual framework to which OASEF
conforms. It includes a theoretical foundation, or a coherent conceptual system,
which consists of of inter-related theories, knowledge, and wisdom that are
shaped or in uenced by individual and social experience, beliefs, philosophies,
and culture [
Moreover, based on the abstract foundation and principles, more concrete
engineering activities are arranged and conducted within a speci c
engineering process, which, in turn, is realised or supported by a range of practical
engineering methodologies. The latter provides speci c and concrete means to
realise the aiming higher level engineering purposes, and to solve practical
human problems in a controllable, repeatable, and e cient fashion [
The overall structure of OASEF is depicted in gure 2. It conforms to the
general framework, and is based on a theoretical foundation rooted in some
multi-disciplinary knowledge, such as general system theory and modern
philosophy of mind. It also incorporates some practical principles by which its
process and methodology are guided or reinforced. These guiding principles, such
as \environment-driven view of service" [
As shown in the gure, OASEF incorporates a exible process model called Organic Aggregation Process (OAP) in the context of service engineering. Figure 3 illustrates its inner structure and inter-relationships at a detailed level. The lines with text besides them represent speci c process activities, whereas the joint point between two lines represents direct correspondence between one activity to another. Moreover, activities are arranged hierarchically in this structure. An activity represented by a single line, starting from point A to point B in the clockwise direction, may comprise a sequence of subordinate activities that are represented by a series of adjacent lines, which also starts from point A and ends with point B clockwise. Superordinate activities are represented by thicker lines and larger-size fonts on the hierarchy. For example, perception comprises subordinate activities of sensation and abstraction. The latter, in turn, comprises system abstraction and general abstraction.
The OAP concepts and process are integrated in OASEF at various level, as illustrated in the middle part of gure 2. The activity of Perception acquires information about the Reality world through Sensation, and forms general or system-speci c abstraction of knowledge. The latter is used by Conception to conduct human intellectual activities in the world of Mind. Speci cally, ConcepWorld Images Rational Images
Capability Images Service Invocation Service Composition Service System Images
Model−driven Method (Eclipse EMF/MOF/AOT) Data, Models, Documents... general abstraction: causation network system abstraction:UML,BPMN... service,process: soaML,BPMN...
Models of various forms Problem situation models Desired End models Service Modeling Service Model, soaML,BPMN...
Process Modeling BPMN, UML Service Implementation BPEL, JAVA...
Runtime environment App Server, Service Engine...
Foundation: System theory, theory of organism and evolution, complexity and chaos theory theory of communicative action, philosophies in epistemology ... tion activities involve Rationalisation, the reasoning and sense-making processes that produce understanding of perceived situations, inferred problems, and desired ends, which are realised by subordinate activities such as Problem Situation and Desired End. Conception also includes high-level design, engineering decision-making, and plan-making sub-processes through Realisation activities, such as Capability Exploration and Capability Design that identify and design desired capabilities in accordance with outcome of Rationalisation. Furthermore Action, conducted in the world of Reality, deal with concrete service systems by nding and invoking existing services, providing new services, or composing composite services. Service Infrastructure provides information of implementation environment such as Enterprise Service Bus (ESB), whereas Service Management provides control and monitoring of services in run-time environments.
Therefore, OASEF derives and manages services using interconnected OAP activities. Achievement of these activities collectively forms speci c images representing uni ed repository of knowledge, which can be referenced, or used as input, by successive activities. For example, the Rationalisation takes World Images from Perception as input. It produces Rational Images that represent rationalised Desired Ends models to deal with identi ed Problem Situation. Realisation takes the Rational Images and produces Capability Images that represent required realisable capabilities. The latter is taken by Action as input and eventually produces SOS that alter the state of the Reality with an aim of improving rationalised problematic situations. 2.2
The conceptual model of OASEF is brought into reality using speci c modelling techniques and model-driven methodologies. Various abstract models, either general purpose modelling language such as UML, or specially designed modelling languages, are used as essential artefacts to facilitate OASEF activities, and to link these activities altogether throughout the engineering life cycle.
OASEF emphasises activities of Abstraction and Rationalisation in the early stage of the life cycle. The objective of Rationalisation modelling is to analysis and justify the rationale and needs of engineering processes toward sensible decisions of actions. Two types of modelling formalism, namely Problem Situation Models (PSM) and Desired End Models (DEM), are designed to capture the crucial higher order cognitive achievement. For example, output of Problem Situation activity is captured in terms of complex inter-relationships between various problems, facts and constraints, and high-level purposes. Together, PSM and DEM provide an important means to explore and represent problematic situations and desired ends that justify successive engineering activities and their produced engineering results.
As an example, the graphical notations of DEM are depicted in gure 4. Coloured rectangles with names inside represent various desired ends elements. They capture the engineering intentions in terms of their concreteness and scope. For example a Goal is more speci c in scope, less abstract to de ne and communicate, and easier to measure, compared with Objective and Ideal. Speci cally, light yellow rectangles represent Ideal, whereas light green and dark green ones represent Objective and Goal respectively. Moreover, white rectangles are used to represent higher order Capabilities that provide contextual meaning and desired value in support of service identi cation and realisation. Relationships among various DEM elements are represented by connecting arrowed lines. For example, an arrowed line can link a Ideal with its subordinate Objective, a Objective with composing Goal, or a Goal with its subordinate one.
Every OASEF activities produce either domain-speci c or general-purpose models that all conform to a uni ed meta-meta-models. Therefore, activities in OASEF can be conducted and inter-connected in a coherent fashion by manipulating and utilising every model in the same way. For example, UML models based on Eclipse Modelling Framework (EMF) are used to capture information and knowledge in Sensation and Abstraction, in the same way PSM and DEM are manipulated. UML activity diagram, soaML, State Machine, and BPMN are used in Realisation similarly. Some models transformation techniques on Eclipse modelling platform are also integrated in OASEF to transform various OASEF models into di erent forms and generate desired system in an automatic or semi-automatic fashion.
Moreover, OASEF provides a mechanism called Epitome to reuse engineering capacities to conduct OAP activities. It is a typical and justi ed means, or capability to achieve some engineering purposes. It is generalised from proven examples that tightly link two OAP activities together, such as a speci c Realisation that is able to achieve predictable and optimised results in reality to improve identi ed problematic situations. Applying an existing Epitome directly produce a previously proven outcome without having to go through the particular activities. For instance, it generate speci c Realisation models that are able to improve the typical problems.
A number of supporting tools are integrated within an integrated supporting tools environment, called IPEOAP. It is developed using Java programming language on top of Eclipse IDE, EMF and modelling platform. A range of graphical model editors are developed to view, create and modify models for various OAP activities, such as PSM and DEM. 3
Two controlled case studies were conducted as proof-of-concept, as oppose to proof-of-performance, which aim to assess whether, in general, its objectives are meet in real world settings. The rst case study is in the context of online travel booking business, a typical scenario used in service community. While the second one is based on Australian First Home Saver Accounts (FHSA) scheme that was introduced by the federal government in 2008 to help residents to purchase their rst homes. This section brie y presents some outcomes of these two case studies to illustrate the main feature, concepts and application of OASEF.
A PSM in the context of online travel booking exempli es the high level analysis of Problem Situation in Rationalisation. It contains many higher level Purposes for online booking, such as \Ease of use", \Quick Responsiveness", \Reliable booking", "Economic price", and \Secure and trust". These PSM elements are generated systematically in accordance with general knowledge captured in general abstraction models within world images. It also contains a range of higher level Problems such as \The booking process takes too long", which violates the \Quick Responsiveness" Purpose. This Problem is also caused by other Problems such as \Too many service providers involved", \Some providers are less e cient than others", \Insu cient network bandwidth", "Diversity in interface and interaction mechanisms", and "The ordering process is too complex" that is caused by \sequential correspondence". Exploration of problems and purposes reveal their nature and relationships and help to identify, understand, and capture important problematic situations.
The PSM also help to systematically generate other models during successive
engineering activities. As an example, since the problem \Online travel booking
takes too long time" is identi ed as a major problem that a ects a higher
priority purpose \Quick Responsiveness", a DEM, depicted by gure 5, is created
to reveal the best desired improvements to address the problem. The
construction process of DEM is guided by, and makes referenced to, elements in the
above PSM. For example, The desired improvement of e ciency during travel
booking process includes a higher level ideal, called \Fast booking", which is
achieved via a number of more speci c Objectives, such as \Service provider
ltering", \Caching", "Simplify booking process", and \Parallel processing",
which, respectively, lters out slow service providers, provides cache to provide
repeated information locally, simpli es the booking processes, and interacts with
service providers in a parallel and asynchronous fashion. The last Objective
contains some lower-level speci c Goal s, such as \Parallel booking" and \Parallel
queries". The identi cation of these improvement and desired ends at di erent
level of details helps to discover the \right" and achievable goals and desired
capabilities to address the important problems.
In OASEF, higher level models such as System Abstraction, PSM, and DEM
are used to derive and capture intellectual achievement as crucial engineering
resources. In the meantime, more concrete design is also captured by models. For
example, UML Activity diagrams are used to describe the internal structure and
behaviour of subordinate business processes in Capability Design, with the aim to
achieve desired ends in DEM. A design of a process to \Close an existing FHSA
account" is systematically derived in accordance with the identi ed problem of
\Cannot manage FSHS account in current bank system" in its PSM. It contains
interconnected required capabilities within an orchestrated structure, such as
\Check Eligibility", \Acquire state information", \Validate Customer Closure
Form", \Fund processing", and \Notify Customer Result". When conducting
Service Composition during Implementation, the above Capability Design model
is automatically transformed into more concrete formalism, such as Business
Process Execution Language(BPEL) models, which can be directly executed in
a run-time environment, such as Apache ODE BPEL engine used in the case
studies. The transformation process from UML Activity Diagrams to BPEL can
utilise various BPEL transformation tools such as the MDD4SOA tools set [
Some observation and assessment of OASEF are made during the design and implementation of proof-of-concept systems in real world settings. Due to a focus on, and methodological support for, higher-level intellectual e orts, OASEF helps to analyse and understand a complex situation, and to systematically derive sensible and rational business decisions, for example, in one of the case study, whether or not provide support for a bank business under speci c legislation environment. The modelling mechanism and tool environment enables developers to concentrate on exploring and understanding higher order business issues including complex situations and desired business ends to be achieve, using models such as PSM and DEM. The graphical tools provided in IPEOAP facilitate the manipulation of these higher-level intellectual e orts, such as design and manipulation of Abstraction, Problem Situation, and Desired End. The use of uni ed EMF modelling meta-model enables interactions and cross-references among various OASEF activities, using either general purpose UML models or specially designed models such as DEM.
In the meantime, \accidental complexity" of underlying implementation technologies is largely hidden during the case study due to capturing and exploitation of engineering capacities by using Epitomes and automatic model transformation techniques. For example, the processes to create database persistence logic, web service provision and invocation, and web user interface are completely captured in various epitomes. Consequently, given a System Abstraction model such as information model for a ight or bank account, OASEF enables 100 percent automatic code generation, which produces systems capable of collecting information from users, persisting data, and providing relevant information service. Experience from the case studies shows that, based on well captured higher order models and technological capacities, generation of concrete systems at the implementation level is relatively fast, and requires little human manipulation.
Some issues and limitations are also revealed by the case studies. Although the proof-of-concept case studies were based on real world settings, the implemented systems are not assessed in real business environment. In fact, since the controlled case studies were designed to demonstrate and evaluate the objectives of OASEF, by the same person who designed the framework, the results are hence less convincing compared with empirical practice by third parties. Moreover, various OASEF activities are neither monitored nor validated in the engineering process, which makes it hard to identify problems when things go wrong. Furthermore, due to its proof-of-concept purpose, these case studies did not cover all aspects of the framework such as sensation and control. It hence requires further work to improve the prototype and conduct thorough empirical evaluation in practice, such as applying evaluation metrics to quantitatively assess the e ectiveness in real world projects, ideally on a benchmark system, and in comparison with other approaches.
Based on the empirical experience from the case studies, OASEF is also
compared with a range of other approaches, which have varying focus, objectives,
and realisation methods. Kohlborn et. al. proposed a consolidated approach that
aims to provide a good business and IT alignment by layering them separately
with certain linkage in between [
Lamparter and Sure also proposed an interdisciplinary methodology that
combines a Web Service engineering method with market engineering and
ontology that aims to coordinate services and customer in a collaborative
environment [
There are also a number of other model-driven approaches to service
engineering [
In this paper, we propose a synthetic approach to service engineering. It incorporates some inter-related elements including a conceptual foundation and guiding principles, a novel process model, a model-driven method, and an integrated supporting environment.
Two proof-of-concept case studies were conducted in real world settings, and are brie y presented in this paper to illustrate some important concepts and techniques. The results show that this approach can be applied in real-world settings to facilitate service engineering and achieve its design goals in general. More speci cally, the following positive results are observed. Firstly, by using higher order models such as PSM and DEM, OASEF enables and promotes coherent higher order intellectual activities, by which e ective services and systems are presupposed. Models in Abstraction, Rationalisation, and Realisation, are captured as important engineering resources that can be located and aggregated together. They hence form important human intellectual assets that provide creative and valuable essence to achieve the \right" and optimised systems. Secondly, activities involving implementation technologies in Action, are also captured as reusable engineering capacities, and are used to automatically realise identi ed desired ends in a relatively easy way. Thirdly, using a model-driven method, OASEF creates an e ective linkage between higher order intellectual e orts and lower level implementation processes, since both type of resources are organised in a uni ed, identi able, reusable fashion. The coherently aggregation of resources and capacities are used altogether to systemically and automatically drive the engineering process to produce desired SOS.
Some important issues and limitations are also revealed during the prototyping and evaluation processing. Corresponding improvements and more thorough evaluation of its practical e ectiveness in real world projects are essential to consider in future research.