Analysis of Mobile Agent Optimization Patterns

Aims: This study extends classification of mobile agent design patterns to involve optimization patterns. We propose two optimization design patterns for mobile agents namely: V-Agent Optimization Pattern and P-Agent Optimization Pattern. The purpose of this paper is to report agents' performance based on mathematical computing model and to support reusability of designs in mobile computing area.
Study Design: This study was developed in collaboration between researchers from computer science department of Amman Arab University, department of computer engineering, department of Network of Applied Science University.
Place and Duration of Study: Samples were implemented in the Department of Computing and Mathematics (Computer Science) and Department of Information System of Amman Arab University for Higher Studies between October 2012 and July 2013.
Methodology: Sample of four master mobile agents that create three slave mobile agents. Master agents are working on set of clients using Aglet alpha release 2.0.5, Tahiti working server and Java Execution Environment (JEE) platform. Slave agents are created by master agents that receive multiple messages from master agents. Master agents send 4000 messages. Consequently, slave agents record message response time in milliseconds. Finally, optimized computed time is computed using the two optimizers: V-Agent Optimizer Design Pattern and P-Agent Optimizer Design Pattern.
Results: Different sample sizes of two data sets are analyzed: seven master mobile agents vs. four slave mobile agents (number of master agents is larger than slave mobile agents, five master mobile agents vs. five slave mobile agents (number of master mobile agents equal to number of slave mobile agents), and four master mobile agents vs. seven slave mobile agents, (number of master mobile agents is less than number of slave mobile agents). Results are based on two main factors: first, disparity of mobile agents in the first data set is computed with variance σ2 is less than 6 and the second data set, variance is larger than 6 between master mobile agents and slave mobile agents, second, number of messages was still fixed in the two data sets of 4000 messages.
Conclusion: finally, it was reported that the optimized computed time for both data sets using P-Agent Optimizer Pattern is less than V-Agent Optimizer Pattern in all cases of different master mobile agents vs. slave mobile agents. It was concluded in this study that P-Agent Optimizer Design Pattern is more efficient, reusable and scalable than V-Agent Optimizer Design Pattern.