Research And Implementation Of Load Balancing In Mass Secure Card System

The Mass Secure Card System designed in this thesis combines the current advanced smartcard technology, network communication techniques and computer science, and forms an intelligent security authentication system. This system implemented the bidirectional authentication between a smartcard of a certain user and that of this system, completion of finance trade in period of validity, and many other smartcard applications. The Mass Secure Card System is working as a smartcards pool at the server end, in order to increase the throughput of the server, and assure the performance of responding large amounts of user requests.Due to the stochastic character of task arriving, and the difference of processing ability among the nodes, some of the nodes will be overloaded after running a while; however, some of the others will be disengaged. On one hand, the overloaded nodes have to finish jobs as soon as possible; on the other hand, those free nodes are somehow a waster of resources. To avoid this situation, we have to design a load balancing algorithm, so to raise the utilization of the entire system, and decrease the mean response time of jobs. The target of load balancing should be minimizing the response time of those tasks.The most challenging part of this thesis is how to design an effective load balancing algorithm, combined with the specific characters of this system. Thus the main task of this thesis is not only to implement the basic functions of The Mass Secure Card System, such as reading and writing IC cards, but also to design an effective load balancing algorithm, so to make the full use of the hardwared parallel processing ability. After summarizing and comparing lots of the current load balancing algorithms, analyzing the features of this system, two new load balancing algorithms were designed, which are Job-Oriented and Card-Oriented Algorithms. Additionally, experiments to quantify the mean response time of each algorithm were carried respectively to compare the performance of the algorithms. The results indicate that these two algorithms have relatively higher efficiency under different system utilization respectively. Besides, experiments are also carried to testify the ability of expand of each algorithm.