Query Processing on OLAP System with Cloud Computing Environment

Abstract

The conventional researches on a distributed On-Line Analytical Processing (OLAP) system have been in hardship to be adapted to real business environment. However, the recent spread of Cloud PaaS (Platform as a Services) provides new chances in the field of a distributed OLAP. OLAP query execution costs many minutes by its enormous data and OLAP query properties. On the other hand, MOLAP has fast responses. But it has a physical space limit to materialize all cells in possible combinations. Therefore, MOLAP is unsuited to analyze large data. In this paper, to provide not only flexibility and expandability of ROLAP but also the speediness of MOLAP, the cloud server architecture is proposed which shares clients’ cube cache by P2P and manages central index on cube data on P2P nodes. In this paper, each node acts as not only a P2P duplicator of aggregated data cubes but also a hybrid server to process queries on sub-cubes. Also, the requests on server are confined to non-aggregated areas and multiple client nodes exchange data simultaneously and asynchronously in the proposed Cloud P2P OLAP. In particular, data are retrieved from physically or logically adjacent nodes. With time series properties, the volume of requested data is minimized and the reuse of past cache data is focused. While central managed P2P has blocked the extendibility, Cloud P2P OLAP guarantees the performance and the extendibility. While Grid OLAP should keep self-distributed system, Cloud use services as much as used by public services. Therefore, Cloud P2P OLAP solves a lot of theoretical limits of conventional distributed OLAP.