Date of Award

12-15-2014

Document Type

Open Access Thesis

Department

Computer Science and Engineering

First Advisor

Csilla Farkas

Abstract

Traditional database concurrency control methods use locking, timestampordering, and optimistic-ordering to achieve DB consistency. However, these approaches are not suitable for long-running Web Service Compositions (WSCs) due to associated performance degradation. Our hypothesis asserts that, using transactional semantic and ordering information, the execution time of a WSC can be reduced, thus allowing the use of traditional database concurrency control methods while avoiding unacceptable performance degradation. Our solution is based on the following approaches: § We model a WSC as WS-BPEL specification, i.e., a partial order of WS transactions. § We allow some of the WS transactions in the WSC to execute in parallel. § We use traditional locking mechanism for WSC to guarantee database consistency. To identify WS transactions that can execute parallel, we analyzed the WS-BPEL specification of the WSCs. The research tasks follow: § Task 1: Identify WS transaction precedence relations § Task 2: Build Parallel Execution Scenarios (PES) § Task 3: Investigate possible further improvement of WSC execution schedule. For Task 3, we propose the following sub-tasks: o Increase the number of WSs executing in parallel, and o Execute concurrently those WSs that have similar execution time In our work we will present our theoretical model and complexity calculation.

Share

COinS