The Research Of Load Balancing Algorithm In Realtime Charging System

The rapid growth of information technology and industry have played an incredible impact to nation's economic and personal life style.With the popularization of smart terminals,3G&4G mobile network and home broadband,the on line billing subject could not be ignored during actions like E-shopping,telecom charge and electric energy telemetering.The request for billing timeliness was enhanced by steps.Take telecom industry for instance,almost all of the telecom billing system had been moved from offline mode to near real time billing system or hot billing system.While the company I was employed have business in worldwide,most of clients of our company already moved to real time billing system or have a plan to make this change.This trend has now become irreversible.Most of the research now are focused on the near real time billing system.Though there are some thesis which take the load balance for real time billing system in cloud computing as subject,those focused on load balance for real time billing system without cloud computing especially with real engineer application were hardly found in both CNKI and IEEE database.So it is worth to dig more in the design and application of load balance utilized in real time billing system.As to the load balance algorithm design in real time billing system,three major popular algorithms development including two static algorithms which are Round Robin and Weighted Round Robin Algorithm and two dynamic algorithms which are Weighted Least Connection Algorithm and Fastest Response Algorithm were selected as research target.The logic and process of the three algorithms applied in ideal load balance system was addressed first.In the real application of real time billing system would add more challenges like more clients,the performance difference across servers,etc.This thesis gave out solutions for all cases above and realize these three algorithms after taking complicated situations into consideration.The key contribution of this thesis was to offer two new load balance algorithms applied in the real time billing system.One is self-defined algorithm to meet users' specific requirements.This new algorithm will explain with several examples like determine on the phone number's parity,assigning specific server to respond at certain time,or assigning specific sever for specific users,etc.The related critical codes was listed also.The other one is distributing algorithm with JGroups.It utilizes the java tool named JGroups to realize communication across different server groups.This one could realize the balance of server load and mutual data copy without needing a complicated configuration for server balance controller.Besides this one could give out a flexible response during abnormal cases,and realize the continuous communication with help from physical sever and memory sever besides avoiding any overload happened in any sever.The load balancing server will response in a short time when the faulty billing server recovers or new billing servers adds to the computer cluster.It will re-assign partitions to archive new load balancing performance.In the end of this thesis,the possibility for distributed load balance algorithm with JGroups used in real time billing system was verified through test data with DCC meter.Through data delay,CPU occupation,memory occupation and wait time of system I/O etc.,it was proved that the distributed load balancing algorithm could realize the expected performance of load balancing.