Research On OLAP-based What-if Analysis Technology On Cloud Platform

In recent years, as the business environment is constantly changing, decision-makers need to capture opportunities faster and more accurate. Mining the value of information from massive amounts of historical data stored in DW takes a crucial role, to enterprise's operations and strategy adjustment. OLAP is an important method for decision support, in OLAP system, what-if analysis is a kind of OLAP technology which is based on historical data analysis of what-if scenarios, it plays an important role in prediction and evaluation for decision-maker. Delta table based what-if analysis is a universal solution, however, the fast implementation what-if update and the execution efficiency OLAP query are the current problem. With the development of massively parallel processing data warehouse, providing efficient parallel query processing and powerful data analysis ability. While cloud computing with high scalability, high computing capacity, resource integration, high fault-tolerant, massive data processing, and low cost, etc. Using massive parallel processing and the advantages of cloud computing is of great significance to solve the performance problem of delta table based what-if analysis.The paper's main works include:(1) Designing and implementing delta table based what-if analysis of cloud services. Firstly, using the Proxmox VE to built lightweight cloud platform, massively parallel processing DW Greenplum to stores data for what-if analysis to provide environmental support; secondly, providing what-if analysis Engine based on Mondrian; finally, realizing what-if analysis of cloud services based on Docker.(2) Designing two scheme to test performance of parallel processing for what-if analysis based on distributed computing, and then putting forward a parallel execution strategy for distributed aggregation function taken a long time for OLAP queries;(3) Analyzing the query performance bottlenecks of delta table based what-if analysis, and in the phase of what-if update, a novel bloom filter optimizing algorithm was proposed, the experimental results show that the performance of what-if view generalization is improved by 32% than traditional join algorithms; in the phase of OLAP query, designing and implementing WFQC which is query caching algorithm based on Redis, our methodology is twice as materialized views based approach.