Analysis, Partitioning and Distributed Execution of XML E-Commerce Transactional Web Services

碩士 === 國立東華大學 === 資訊工程學系 === 93 === Web services have brought a new age for E-commerce applications. Enterprises can reorganize their inner resources into Web services based on service-oriented architecture. This leads to a new approach for cross-enterprise cooperation and operation. In such case, t...

Full description

Bibliographic Details
Main Authors: Jin-Hou Lin, 林俊豪
Other Authors: Shiow-Yang Wu
Format: Others
Language:zh-TW
Published: 2005
Online Access:http://ndltd.ncl.edu.tw/handle/09547421378217112061
Description
Summary:碩士 === 國立東華大學 === 資訊工程學系 === 93 === Web services have brought a new age for E-commerce applications. Enterprises can reorganize their inner resources into Web services based on service-oriented architecture. This leads to a new approach for cross-enterprise cooperation and operation. In such case, transactional support in Web service composition will become one of the most important issues in this architecture. An integrated solution for flexible Web service composition, distributed service execution, and transaction management is therefore in strong demand. We propose a comprehensive architecture for cross-enterprise business modeling, Web services composition, and distributed transactional execution. We use an XML-based modeling language, named AC-Diagram, as our Web services composition tool to automatically generate cross-enterprise workflows from visual diagrams. We also design a flow partitioning algorithm to partition a complex workflow into modular subflows. Based on a detail analysis of the formal semantics of AC-Diagram, the partitioning algorithm retains all the causal ordering of the original workflow while exploiting as much as possible the available parallelism. The resulting subflows can then be dispatched to our flow execution engines which are fully distributed Web services capable of executing the subflows and coordinating with each others for the successful completion of the entire business process. To ensure transactional execution, three algorithms, namely flow execution, rollback and post-processing algorithms, are proposed to control and coordinate the execution engines. When the computing resources are limited, we employ a novel approach based on graph theory to reorganize subflows into clusters such that the computational load is balanced and the communication cost is minimized. Experimentation results demonstrate that our approach can effectively exploit network computing resources for executing complex cross-enterprise business processes with transactional support and very low overhead.