Data Driven Adaptation of Heterogeneous Service Oriented Processes

Within the currently forming pervasive computing environment services and information sources thrive. Instantiations of the service oriented computing paradigm e.g. Web, Peer-to-Peer (P2P) and Grid services are continuously emerging, whilst information can be collected from several information sources e.g. materialisations of the Web 2.0 and Web 3.0 trends, Social Networking apps and Sensor Networks. Within this context the development of adaptable service oriented processes utilising heterogeneous services, in addition to available information is an emerging trend. This paper presents an approach and an enabling architecture that leverage the provision of data-driven, adaptable, heterogeneous service processes. Core within the proposed architecture is a set of interacting components that accommodate the acquisition of information, the execution of service chains and their adaptation based on collected information. 1 Dissertation Summary Our era has been marked by a shift in the way of thinking and acting across many domains such as business, science and community. Cornerstone in this new way of thinking is the notion of service. In spite of the lack of a commonly accepted definition a service can be conceived as a function that is offered by someone and may be used by anyone else, or as Douglas Barry sets it more formally: a function that is well-defined, self-contained, and does not depend on the context or state of other services [1]. This shift in thinking and acting has an effect on software systems and the way they are developed. Service Oriented Computing (SOC) focuses on the use of services as system constituent parts. It is regarded as an evolution to the Component-Based development and distributed object oriented computing, and has been widely accepted as the current and future trend in distributed system development. Among its goals is to promote the loose coupling and flexible integration of the system parts in a far better way than component and object oriented technologies do. ? Dissertation Advisor: Aphrodite Tsalgatidou, Associate Professor 1.1 Service Oriented Process Adaptation Service Oriented Processes (SOPs) constitute an indicative materialization of the loose coupling notion of SOC, as they rely on the lenient integration of comprising services. They are normally defined in higher-level languages, e.g. BPMN [2], WS-BPEL [3], and provide descriptions of coordinated flows of constituent (atomic or compound) services. One of their prime characteristics that has contributed to their proliferated usage is their easy execution by contemporary orchestration engines, e.g. Apache ODE . Even though, SOP development is regarded as an evolution of Enterprise Application Integration approaches [4], it is also a new paradigm that is referred as Mega-Programming [5]. An inherently assigned characteristic, stemming from this consideration, is that processes are expected to be long-running and stateful as well as they may involve the interaction with stateless or stateful services. In the frame of our approach, SOPs are regarded as systems, which operate within a specific environment, and perform pre-specified activities, via the use of external services, in an orderly manner producing and/or consuming related information during their execution. Nonetheless, in the currently forming Pervasive Computing environment, resources such as services and information sources emerge with an increasing pace. The Service-Oriented Computing (SOC) model has been instantiated by several distinct paradigms, e.g. Web, P2P, OGC services, and a plethora of service instances are emerging every day. In addition, the emerging Sensor Web [6], and the materializations of the Web 2.0 [6] and Web 3.0 [7] paradigms, e.g. Social Networking applications, provide new types of information sources. In this context, rendering SOPs able to tap onto these resources is becoming a necessity rather than an option. These resources could be used for the adaptation of SOPs with the goal to optimize their execution. Hence, processes should be flexible enough in order to facilitate the use of services that have not been identified at design time, irrespectively of their type, and the use of information that may stem from emerging sources. This notion of process adaptation differs from contemporary definitions, e.g. Cassati et al [8], in that process adaptation should also support process optimization. Hence, the benefits acquired by process adaptation should be measurable in terms of specific indicators (criteria). The list of criteria that can be used for the optimization of SOPs may comprise cost, execution time, throughput, etc. In the scope of our research we argue that process adaptation should be used for the optimization of the process execution time. An approach to achieve this optimization is via the reduction of unnecessary process activities. Considering the properties of SOPs, i.e. long-running activities, it is plausible to expect that the reduction of unnecessary process activities can lead to smaller execution times and as a result to higher throughput. 1 http://ode.apache.org/