Performance and reliability improvement of the call detail record processing system: A case study from a telecommunication enterprise

Several data processing systems in a telecommunication enterprise rely on batch scheduling. The batch systems usually contain a large amount of data and require complex scheduling process. Using native job launchers like Cron cannot handle these requirements and will likely degrade performance and reliability of the system. In this paper we propose to custom-design a specific job scheduler with dynamic priority assignment for the Call Detail Record (CDR) processing system at a telecommunication enterprise. Our scheduler is designed based on the Sourceforge's Open Source Job Scheduler (JS) software with the rate-monotonic (RM) scheduling algorithm. The algorithm is used for on-time job completion purpose. We then evaluate the results in both the perspective of performance and reliability improvement. We also compare the results with the native job launchers. The performance metrics used are the speed of data processing and the number of over-period job. The reliability metric used is the completeness of data in the system.