Login | Register

Full Solution Indexing and Efficient Compressed Graph Representation for Web Service Composition


Full Solution Indexing and Efficient Compressed Graph Representation for Web Service Composition

Li, Jing (2016) Full Solution Indexing and Efficient Compressed Graph Representation for Web Service Composition. PhD thesis, Concordia University.

[thumbnail of Li_PhD_S2017.pdf]
Text (application/pdf)
Li_PhD_S2017.pdf - Accepted Version
Available under License Spectrum Terms of Access.


Service-oriented computing enhances business scalability and flexibility; providers who expect to benefit from it may bring explosive growth of web services. Searching an optimal composition solution with both functional and non-functional requirements is a computationally demanding problem: the time and space requirements may be infeasible due to the high number of available services. In this thesis, we study QoS-aware service composition problems which satisfy functional requirements as well as non-functional requirements. We use optimization algorithms to enhance accuracy of our searching algorithms.

In the first approach, we propose a database-based approach to search a service composition solution. Current in-memory methods are limited by expensive and volatile physical memory, to deal with this problem, we want to use the large space available in relational database on persistence disk. In our database-based approach, all possible service combinations are generated beforehand and stored in a relational database. When a user request comes, SQL queries are composed to search in the database and K best solutions are returned. We test the performance of the proposed approach with a service challenge data set; experiment results demonstrate that this approach can always successfully find top-K valid solutions.We offer three main contributions in this approach. First, we overcome the disadvantages of in-memory composition algorithms, such as volatile and expensive, and provide a solution suitable to cloud environments. Second, we fetch top-K solutions in case the optimal solution is not available as backup solutions to the user. Third, compared with other pre-computing composition methods, we use a single SQL query: there is no need to eliminate spurious services iteratively.

Then, we propose the application of a skyline operator to reduce the search space and improve the scalability. Skyline analysis returns all of the elements that are not dominated by another element. We use skyline analysis to find a set of candidate services referred to as "skyline services", therefore, less competitive services are reduced. This allows us to find a solution for a large composition problem with less storage and increased speed.

In reality, different users may have same requests, we are motivated to pick some popular requests and generate paths for fast delivery. These paths are stored in a separate table of the relational database. When a user request comes, we first search to find a nearly ready-made solution. Only as a last resort do we search the table with whole paths to find a solution.

Finally, to deal with the problem that the search space may explore, we apply a compressed data structure to represent the service composition graph. The goal is to allow algorithms running in in-memory over larger graphs. In this approach, we use compact K2-trees to represent the service composition graph. When a user request comes, we search the K2-tree for a satisfactory solution. We use an array to store values in the last level of the compact tree, which represents relationships between services and concepts. In our algorithms, we find services' inputs (resp. outputs) by locating elements in this array directly, therefore, decompressing the graph is unnecessary. To the best of our knowledge, our work is the first attempt to consider compact structure in solving web service composition problems. Experiment results demonstrate that this approach takes less space and has good scalability when handling a large number of web services.

We provide different approaches to search a solution for the user. If the user want to find an optimal solution with fewer services, he may use the database-based approach to search for a solution. If the user want to get a solution in a short time, he may choose the in-memory approach.

Divisions:Concordia University > Gina Cody School of Engineering and Computer Science > Computer Science and Software Engineering
Item Type:Thesis (PhD)
Authors:Li, Jing
Institution:Concordia University
Degree Name:Ph. D.
Program:Computer Science
Date:October 2016
Thesis Supervisor(s):Yan, Yuhong and Lemire, Daniel
ID Code:981954
Deposited By: JING LI
Deposited On:31 May 2017 18:07
Last Modified:18 Jan 2018 17:54
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Downloads per month over past year

Research related to the current document (at the CORE website)
- Research related to the current document (at the CORE website)
Back to top Back to top