Development Of Common Subscriber Data Base And The Improvement Of The Scheduling Algorithms

In the mobile communication systems, HLR is in the highest level of the network that responsible for managing the database of mobile users, information of the registered data and its location information, providing routing information for mobile core network. AUC of the PLMN is the security management unit of the mobile communications system which stores the necessary information for protecting the mobile users'communication. In general, AUC and HLR are equipped together in an unified entity (HLR/AUC).With the rapid development of mobile communications and the sharp increase of the mobile user data, there have been higher requirements of safety, reliability of real-time access and scalability for the HLR/AUC system. The unique integrated systems architecture will not be able to meet the rapid development of mobile communications industry. So the separate database system based on a common framework pushed the prices sky.The main work of this paper:1. Participated in e the software and hardware design and development of CSDB (Common Subscriber Data Base) systems based on ATCA technology.2. Participated in the design and development of GLOAD component. The system has been applied in the Vodafone UK, China Union and so on.3. Designed a special algorithm of Intelligence Segmented LFU based on the design and development of CSDB systems..4. Proved that the algorithm will improve about 5% than the original algorithm used in LFU algorithm. This paper designed and developed the framework of software/hardware, redundancy handling of hardware/network of CSDB system which based on ATCA architecture. And then designed and developed the OVERLOAD component of the system. This part is designed for monitoring the system load conditions. OVERLOAD component composed by two parts:GLOAD and SLOAD.GLOAD is equivalent to the monitoring service center. The main task of it is that unpacking, sorting, extracting, analysis the monitoring data, control data and load-balancing data send by the SLOAD. GLOAD collects all data of the transmission, CPU occupancy rate and resource utilization, and compare with the defined threshold. The results will be passed to each of SLOAD. Related SLOAD will take the action for the results.SLOAD is equivalent to the monitoring system of client-side with on-line monitoring, alarm classification functions and so on. SLOAD load configuration based on the manual settings and systems automatically adjust. In RT Blade, specific data are handled (and/or compute) by a local agent named SLOAD. The local agents of all the stations involved in the regulation function communicate periodically their results to GLOAD.The transmission quantity of information between GLOAD and SLOAD are not so much, so session model is used to be the interface between the GLOAD and SLOAD. The GLOAD received data containing calculations the results sent by the SLOAD and sends back the status of the entire platform to SLOAD.System indicates the relationship between the application and SLOAD through the application ID. GLOAD send the message to the SLOAD regularly.Finally, the paper designed a set of improved algorithms for the cache system through the research and the development of the system. Based on the Index Server of this system, the user applications data and the storage data has been separated, and then there has been a problem of how to read the user data. In this system, first read out the data of user access from DB, and then stored them in the Index Server on FE, and FE will visit them. The framework of the design make efficient paging algorithm to improve its performance become more urgent.In traditional cache scheduling algorithm, there are FIFO (First In First Out), LRU (Least Recently Used) and LFU (Lowest Frequency of Use). For mobile users, the main access characteristics are embodied in the user record access frequency, so LFU algorithm is the best choice. However, the user's access frequency also reflects cyclical fluctuations in a short time. It is easy to transferred data out of cache which have the potential of high-frequency access in the future, while its data is not a good trend of convergence in the user's access. Therefore, according to cyclical nature of mobile users access, this paper presents a new intelligent LFU algorithm to access the data based on the historical analysis, to predict the future of user data access probability curve, reducing the frequency of user data eliminated and system resource consumption, and system stability problems decline caused by the system resources occupancy rate is too high.Intelligence Segmented LFU algorithm had sub-processing based on the LFU, the cache divided into three levels of high-Middle-low access areas. Each size is variable, while each segment using different replacement algorithm to adapt to different levels of access frequency of different characteristics. Same as the LFU's counter function, whenever a user record has been accessed, the corresponding access to counter auto-plus-one, while the rules based on three visits to adjust the size of the area to improve the cache hit ratio.When the first priority of the user access reached the boundary value, the user will be moved to higher priority areas at the next time which indicates the level of activity that the user has reached a higher level. When the active users reached the highest priority, the user only exchange with the previous and cannot raise the priority level, until reach the most front-end. For the high frequency access users, the algorithm could ensure it quickly move to higher priority, and slow phase-out, given sufficient time to prepare for revival of silence, while the post-resurrection quickly restored to a high priority. For random access users, its level promotions depend on that its reach a certain number of visits (boundary value), while the phase-out but also in a period of time (a new user entry) is not being accessed. This can effectively to prevent jitter. Because its priority processing in the implementation of the priority areas, and it could prevent invalid user to stay too long, the three priority areas to ensure access frequency of different users with different buffer levels of the stay.Experiments show that the algorithm is able to enhance the efficiency of 5% than LFU algorithms for 1 million data access that obey normal distribution. For the system cyclical fluctuations, Intelligence Segmented LFU algorithm is able to convergence the predicted values to the actual value during three visits cycle around. That could make sure the stability of the algorithm.