Improvement of a big data statistic system : using multi-level mid-tier caching

This work presents the improvement and optimization of a statistic system containing big data. Both the performance and the visual presentation of the system are addressed but the main focus is against the performance part of the problem.Baseline benchmarks of the performance and analysis regarding the visual presentation of the system are conducted. Multiple solutions are discussed, evaluated and critically analyzed to find the right key for this problem.The technique of implementing a Multi-Level Mid-Tier cache with an "Update on Event" trigger is proposed as a solution to the performance issues in the system. By implementing a hierarchical structure to a mid-tier cache, the idea is to provide a highly dynamic, configu-rable and responsive system without compromises. The Multi-Level Mid-Tier cache is in this implementation consisting of two levels; one static and one dynamic.The creation of a multi-platform compatible user interface is proposed as a solution to the visual presentation. This is done by revising the front end of the system by implementing API; s such as JQMobile and DataTables.Benchmarks of the improved system are performed. Both the loading time of the system and the derivative of the increase of loading time are tested in the same way as for the base-line results. Comparing these results with the baseline results lead towards the conclusion which confirms the sterling functionality of the Multi-Level Mid-Tier Cache.The Multi-Level Mid-Tier cache eliminated the performance problems of the MyMo statis-tic system. The loading time at 10 000 reports is decreased by 96,45% and the derivative at the same point is decreased by 98,40% without the any compromises to the accessible data.