Year

2014

Season

Spring

Paper Type

Master's Thesis

College

College of Computing, Engineering & Construction

Degree Name

Master of Science in Computer and Information Sciences (MS)

Department

Computing

NACO controlled Corporate Body

University of North Florida. School of Computing

First Advisor

Dr. Karthikeyan Umapathy

Second Advisor

Dr. Asai Asaithambi

Third Advisor

Dr. Lakshmi Goel

Department Chair

Dr. Asai Asaithambi

College Dean

Dr. Mark A. Tumeo

Abstract

Business transactions (a.k.a., business conversations) are series of message exchanges that occur between software applications coordinating to achieve a business objective. Web service has been proven to be a promising technology in supporting business transactions. Business transaction can either be long-running or short-lived. A transaction whether in a database or web service paradigm consists of an “all-or-nothing” property. A transaction could either succeed or fail. Web Service Atomic Transactions (WS-AT) is a specification that currently supports Two-Phase Commit (2PC) protocol in a short-lived transaction. WS-AT is developed by OASIS–a standards development organization. However, not all business process scenarios require a 2PC, in that case, just a One-Phase Commit (1PC) would be sufficient. But unfortunately, WS-AT currently does not support 1PC optimization.

The ideal scenario where 1PC can be used instead of 2PC is when there is only a single participant. Short-lived transactions involving only one participant can commit without requiring initial “prepare” phase. Thus, there is no overhead to check whether the participant is prepared to either commit or rollback. This research focuses on designing a mechanism that can add 1PC support in WS-AT. The technical implementation of this mechanism is developed by using JBoss Transaction API. As a part of this thesis, 1PC mechanism for a single participant scenario was implemented. This mechanism optimizes the web service transaction process in terms of overhead and performance in terms of execution time. The technical implementation solution for 1PC mechanism was evaluated using three different business process scenarios in a controlled experiment as a presence or absence test. Evaluation results show that 1PC mechanism has a lower mean for execution time and performed significantly better than 2PC mechanism. Based on the contributions made by this thesis, we recommend OASIS to consider including 1PC mechanism as a part of the WS-AT specification.

Share

COinS