Improved rsync algorithm to minimize communication cost using multi-agent systems for synchronization in multi-learning management systems

Remote file synchronization (RFS) is based on updating the outdated version of a file that resides on one machine to be similar to the new version of the updated file in another machine at a minimum computation time (cost). The problem of rsync algorithm synchronization process is that rsync trie...

Full description

Saved in:
Bibliographic Details
Main Author: Mwinyi, Amir Kombo
Format: Thesis
Language:English
Published: 2017
Subjects:
Online Access:http://psasir.upm.edu.my/id/eprint/71209/1/FK%202017%2073%20-%20IR.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Remote file synchronization (RFS) is based on updating the outdated version of a file that resides on one machine to be similar to the new version of the updated file in another machine at a minimum computation time (cost). The problem of rsync algorithm synchronization process is that rsync tries to check folders and files one by one, which takes long time to synchronize them. Therefore, the aim of this study is to minimize the computation time during RFS by improving the standard rsync algorithm. Previously, several algorithms and techniques have been proposed for partial file synchronization but many of them were based on controlling the block size, checksums, and delta compression of the matched blocks, to solve the searching problem of the rsync algorithm. This study proposed several techniques to improve rsync (irsync) algorithm in order to reduce computation time during RFS, by encompasses a Multi-Agent system (MAS) framework. This algorithm involves several agents, such as: initiator, sense_agent (SA), log_agent (LA), and search_agent (SeA). These types of agents have different capabilities, actions, and efficiency to the irsync algorithm in file synchronization. The study proposed MAS framework in the Learning Management System (LMS) that involves the transfer of data from one machine to another. To meet this requirement, a new Multi LMS (MLMS) model using Sharable Content Object Reference Model (SCORM) specifications to share learning materials among different higher learning institutions (HLIs) has been presented. This model enhances the interoperability and collaboration of HLIs in terms of synchronization and sharing of learning contents. To evaluate the computation time of the new techniques, standard datasets, which include two versions of source codes emacs-19.28 with emacs-19.29, and gcc-4.8.1 with gcc-4.8.2, were used. The experimental results show that the improved rsync (irsync) algorithm yields a better performance against two previous algorithms, rsync algorithm and hierarchical folder synchronization algorithm (HFSA) in terms of reducing computation time and improve synchronization response time. Therefore, the MAS framework was performed and the reduction of computation time was obtained by 19.86% compared to 42.25% of standard rsync. The results also indicated that reducing the searching time could enhance the irsync algorithm responsiveness time by 32.07% compared to 67.93% of the standard rsync. The integration of the proposed MLMS model with irsync algorithm was further tested through a prototype with MAS and show significant improvement over the cloud synchronization system which based on CDMI technology. These two systems were evaluated in terms of synchronization rate. The results revealed that the MLMS system with irsync (MAS) outperformed the cloud system.