Service-based information systems are considered a cornerstone in architecting modern enterprise applications. Service-oriented architecture (SOA) is designed to enable dynamic integration of heterogeneous application elements, and thus improve enterprise agility. This is achieved by publishing reusable services on a common registry or enterprise service bus to become available to users who will request and invoke them according to their business needs. In this paper it is argued that service-based information systems are different from componentbased systems. SOA relies on the concept of contracting services to become invoked by users through service matching and binding, beside a middleware interface that integrates heterogeneous components (as supported by component-based architecture principles). In order to enable service evolution it is suggested to conceptualize service-based information systems as a complex adaptive system. Complexity science seeks to theorize the phenomenon of emergence of new properties and the spontaneous creation of new order, and thus provides elements to realize adaptability and evolution. By mapping complexity principles to SOA features factors that would facilitate information service evolution can be derived. It is concluded that in order to enable sustainable evolution of information services controlling factors, such as contracting, licensing, provenance, reliability and sustainability are paramount beside component-based development principles that include reusability, loose coupling, inter-operability, scalability and platform-independence. New programming discipline practices are also concluded, that include service choreography model, BPEL specifications, meta-data specifications, SLA specifications, service semantics and service tests. These ensure fast development balanced with discipline and quality, beside extensive collaboration and interactions facilitated by conventional agile development practices like JAD sessions and prototyping.
According to
SOAP or REST, Web protocol for communicating service calls and messaging over HTTP or SMTP/email protocol UDDI (Universal Description, Discovery, and Integration), a universal application programming interface (API) that facilitates service registration and searching WSDL (Web Service Description Language) for service description that include information on data structure types returned or invoked by the service, transport protocol used, physical service endpoint, etc. (see Figure 1 for more elaborations) 4) Bind/Invoke
SOAP Service Requestor
Service Provider 3) Reply WSDL 2) Request
Service Registry
UDDI 1) Register
WSDL
Business processes typically encompass multiple service invocations and in this regard
WSBPEL (Web Services Business Process Execution Language) introduced by OASIS describes
language syntax to compose and execute Web services. These include operation commands, such
as invoke, receive, reply, wait, assign, throw and terminate, as well as control flow commands
that include sequence, switch, pick, while, flow and link (www.oasis-open.org). Once a system
architect finalizes the required service composition using a BPEL editor, the relevant BPEL file
is generated. The service can then be made available by storing the respective WSDL file in
UDDI. When the service is invoked by a user the BPEL engine reads the stored BPEL
specifications in order to execute the sequence of required services
According to
In addition,
Reflecting on a real-case technical practice of SOA, a Learning Resource Recommendation
System for universities,
Similarly,
According to
Complexity science seeks to explain the process of self-organisation, emergence of new
properties and the spontaneous creation of new order. CAS theory originated in the natural
sciences and articulates how interacting agents in systems adapt and co-evolve over time in
creative and spontaneous ways
Complexity principles emphasize that emergence of properties and creation of new order are not
explicable from a purely reductionist viewpoint, but the whole is greater than the sum of the
parts (Kaufman, 1993). This means the focus of attention shifted from understanding the parts
or entities of which the whole was composed to the interaction of subsystems (agents) to form a
system. The emergence of order from heterogeneous local interactions is formed when feedback
from environment and interacting agents informs the circular dynamics of the system. Thus
local interactions will influence the formation of global structures and the stability of their
mutual reproduction(
Feedback
Feedback
Regularities
Interactions
Agents
While a uniform description or interpretation of CAS is still not provided, several key aspects
that characterise CAS were suggested in the literature. After reviewing several CAS literature
CAS consists of a large number of components that undergo continuous change processes
and re-arrangements, that in turn will define new identity of the system
CASs are made up of heterogeneous agents (Holland, 1995); each agent is different from
the others (diversity), and its performance depends on the other agents and the system
itself
Tendency to adapt to a particular situation depends on the context and the influence of
environment; this will determine the possibilities and alternatives available for change
Structural coupling emphasizes the analysis of systems and their evolution in terms of
their form, structure & degree of interconnectivity (
In order to harness change with no anarchy CAS strives to maintain a balance between
the completely ordered, “frozen” regime and the completely disordered, chaotic regime,
which is known as operating on “edge of chaos” (Mc
Nonlinearity principle emphasizes that emergence of properties and new order are not explicable from a purely reductionist viewpoint, but the whole is greater than the sum of the parts (Kaufman, 1993). (P7) Interactions:
Agents in a CAS undergo constant interactions, both autonomously and with their
environment
Inter-relations between the system parts result in feedback loops where components in the
output stage inform components in the input stage
As a result of feedback loops adjustment in CAS takes place through the learning
experience exhibited by its agents. This will change the agents’ effect to realise the
required outcome
The behavior of the system in one period of time feeds back and informs to determine
behavior in the next time; this gives the system ‘historical dimension’
Agents rarely are partitioned into non-overlapping groups; they rather participate in
multiple neighborhoods/undertakings simultaneously, where their various activities
coevolve
There is great interest in understanding modern organizations and systems in terms of theories of
complexity
In order to operationalize CAS principles,
It is found that social construction elements, such as communication, collaboration, interaction, etc. are argued to be critical drivers of human empowerment and self-organisation, whereas mechanistic, adaptive dynamics like flexibility, short-term orientation, small scale approaches, simplicity and rapidity will ensure fast response and quick adaptation to the problem situation. However, evolutionary properties cannot be fully realised without the necessary enabling infrastructure that will allow the dynamics of emergence to become effective. Also appropriate control mechanisms, such as feedback, reflection, learning, discipline frameworks and methodologies for management and operation, as well as embracing quality control factors need to be in place in order to ensure emergence to happen without descent into anarchy (edge of chaos). The elements or factors in each category have been identified and related in a framework, to help understand and analyse the phenomenon of emergence and facilitating evolutionary properties in social organisations in general, and information systems development in particular (Table 1).
Communication Collaboration Interaction Short-term orientation Small-scale Rapidity Flexibility Simplicity Management style Culture Reflection Learning Organisation structure Technical architecture Feedback Continuous re-adjustment Quality controls Discipline in programming Minimal development methods
By mapping factors facilitating ISD evolution as implied by CAS (Table 1) to elements and features of service-based information systems development, factors of service systems evolution can be derived. In previous sections specific characteristics of SOA were outlined as compared to component-based systems, along with issues in modelling SOA and development methodologies targeted at successful development, operation and governance of SOA. In doing so, several key elements characterizing SOA implementation have been identified (represented in italic). Mapping such elements to different aspects of the evolutionary process (dynamics, enabling infrastructure, controls) yields the suggested framework for service-based information system evolution represented in Table 2.
For the enabling infrastructure factors like reusable components, loose coupling, interoperability (based on WSDL and application programming interfaces/APIs), platform independence, enterprise service bus and linked-open registries (repositories) have been identified under technical architecture. Under organization structure elements such as enterprise-wide consolidation, streamlining cross departmental processes, migration of legacy processes, etc. For the dynamics of service evolution attributes like autonomous, encapsulation (service discovery & conceptualization), abstraction, stateless, meaningful and standardized to business (Business/IT alignment), designed for contracting, service matching, linking and binding are suggested. Rapidity is also important in SOA implementations in order to cope with the rapid change in current business environments, and increase the enterprise competitive advantage. Collaborations and interactions are also paramount represented in interactions with the vendors, users and between the different departments (for example during JAD sessions). Controlling factors for service-based information systems evolution is different from componentbased development, due to the fact that SOA relies on the concept of contracting and brokering. Therefore, quality control factors such as licensing, provenance, reliability and sustainability become necessary to govern the process of contracting and service matching & binding. This is because a licence should be provided before contracting, as well as a provenance process is required to trace back sources of data for correctness, and a process to ensure that the link to be invoked is reliable (not to disappear or shut down) and sustainable, i.e. other relevant links are provided in case of shut down of the primary link. This is beside the usual quality control factors for component-based development that include security and redundancy removal. In order to ensure the successful implementation of SOAs several development methodologies/approaches have been suggested in literature as discussed before. These also provide controlling elements for SOA evolution, as they govern the process of service development and evolution life cycle. But these approaches need to be not restrictive in order to facilitate rapidity and quick adaptation; i.e. provide minimal guiding instructions without being cumbersome as supported by CAS principles. Agile development rationale to emphasize light weight development is rush into coding but governed with discipline in programming that will ensure quality of programming outcome. In case of SOA new programming disciplines can be concluded; service choreography model, BPEL specifications, meta-data specifications, SLA specifications, service semantics and service tests.
According to
In this regard,
Collaboration & Interaction for SOA ο Cross-company ο IT vendors ο Users ο JAD (Joint Application
Design) sessions Flexibility for SOA ο Encapsulation &
Abstraction (Service discovery, conceptualization & abstractions) ο Meaningful and standardized to business (Business/IT alignment) ο Stateless ο Designed for contracting, service matching, linking and binding ο Autonomous Rapidity for SOA
ο Faster time to service Short-term orientation, Small-scale & Simplicity for SOA They have not been addressed directly in current literature.
Short-term orientation and small scale development counteracts enterprise integration principles, but still need to be embraced in a balanced way due to that fact of rapid technological obsolescence. ο Management style & Culture for SOA ο ο
SOA Best Practices SOA Governance Policies SOA Metrics & Maturity
Measurement Evolutionary Technical Architecture for SOA ο Reusable components ο Loose coupling ο Interoperability (based on WSDL and application programming interfaces/APIs) ο Platform independence ο SOA topology &
Enterprise service bus ο Linked-open registries
(repositories) Organization structure for SOA ο Enterprise-wide
consolidation ο Cross department process
streamlining ο Migration of legacy
processes
Minimal development
Simplicity is an important agile methods for SOA
principle that needs to be more ο SOAD
exploited in SOA (
Modelling
Framework
Model for SOA
Continuous refection & governance Feedback & Continuous readjustment for SOA ο Extensive
prototyping ο Continuous
improvement Quality controls for SOA ο Licensing ο Provenance ο Reliability ο Sustainability ο Security ο Remove redundancy Discipline in programming for SOA ο ο ο ο ο ο
Service choreography
model
BPEL specifications
Enterprise
Integration & meta
data specifications
Service registry and
semantics
Service tests
Review of SLA for
semantic correctness
Complex adaptive systems theory conceptualizes the phenomenon of emergence,
selforganization and spontaneous creation of order. Several generic complexity principles have been
suggested in literature, such as diversity, large number of agents, interconnectivity, interactions,
feedback, edge of chaos, etc. that refer to evolutionary characteristics. In this paper it is
suggested to conceptualize service-based information systems as a complex adaptive system and
in that way derive factors that would facilitate information system service evolution.
After surveying several elements related to service-based information system development, such
as specific characteristics of SOA as compared to component-based development, issues in
modeling SOA, as well as development approaches suggested in literature to govern service life
cycle and evolution, it is concluded that service-based information systems are different from
component-based systems. As implied by
By mapping service-based information system development elements to CAS principles several factors that would facilitate evolutionary properties of SOA have been derived. It is concluded that in order to enable sustainable evolution of information system services licensing, provenance, reliability and sustainability are paramount beside object-oriented and componentbased development principles that include reusability, loose coupling, inter-operability, scalability and platform-independence. New programming discipline practices are also concluded, such as service choreography model, BPEL specifications, meta-data specifications, SLA specifications, service semantics and service tests. JAD and prototyping are also appropriate for SOA development, BUT more attention should be paid to the architecture design as implied by RUP, and also maintenance and governance process is of great importance, as SOAs have more longevity and deployment element than simple information systems that are small scale and short-term oriented.
It is suggested that the introduced framework guides system and IT architects, as well as management teams with factors or strategies that will enhance service-based information system evolution supported by CAS principles. These are preliminary findings and future work will focus on a more detailed mapping of the different factors, as well as application on a real-case study in order to evaluate the proposed framework and put forth metrics for SOA evolution based on identified factors.