Rescuers

: a framework for reliable web services composition

Abstract

Web services composition promises to build new value-added services by aggregating several existing services in an efficient and cost-effective way. As web services are deployed on the unreliable Internet, and as they are often long running, loosely coupled, and cross administrative boundaries, failures are expected to happen frequently during their execution. Considering the fact that a composite service would invoke a number of web services to fulfill its functionality, how to deliver reliable web services composition over unreliable web services therefore becomes a challenging and significant problem. In this dissertation, we present RESCUERS, a framework proposed to address the problem of reliable web services composition by incorporating various techniques, including exception handling, transaction, service scheduling, and service selection. The main research works and contributions of this dissertation include the following: 1. We design a service specification module which enhances the fault tolerance of web services composition while keeping the service design simply and easy. We first propose EXTRA, a hybrid fault tolerant mechanism that combines exception handling and transaction techniques. EXTRA defines a set of high level exception handling strategies as well as a termination protocol. Then, we devise an exception handling strategy builder to assist service designers to specify fault handling logic efficiently by using declarative ECA rules that are built on top of the exception handling strategies. We also propose a logic verifier to automatically verify the correctness of composite service from a transactional point of view. Further, we devise a logic integrator to transform the fault handling logic described by ECA rules into standard WS-BPEL codes. We demonstrate the viability of the service specification module through experiments. 2. We propose a service scheduling module which ensures the execution atomicity of composite services (with minimum cost) even if service compensation has temporal and cost constraints. Specifically, we devise a time-aware scheduling algorithm, which adopts the critical path method (CPM) to compute for scheduling time information on a time composition graph that is adapted from timed workflow graph. We analyze the expected compensation cost of composite services, and then devise several algorithms to generate optimal or near-optimal schedules by using backtracking and genetic algorithm. A dynamic scheduling mechanism is presented to deal with the variations of execution durations of services, which complements the proposed scheduling algorithms. We benchmark our algorithms by simulations and the experimental results show that they are effective. 3. We devise a service selection module which reduces the possibilities of runtime faults by choosing highly reliable services and keeps down service usage fee at the same time. For this purpose, we propose a new concept, profit of composition, to measure the contributions of service QoS to web services composition from an economic point of view. We give a business model and a QoS model for web services, based on which we elaborately explain how to assess the profit of composition, and also present a profit driven selection approach, whose performance is validated by a variety of experiments.

Date of Award	2 Oct 2009
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Qing LI (Supervisor)