Horizontal cloud database partitioning with data mining techniques

While horizontal partitioning of traditional relational databases is well studied, there is in comparison very little in the literature about the partitioning of NoSQL databases, especially those that are cloud based, such as Amazon´s SimpleDB. In order to optimize the response time of a web application using a cloud based database, it is possible to partition the database for storage on local regional servers. In this work, data mining and cluster analysis of database logs is used in order to partition, thus essentially caching, data on local servers. This work compares the average response times of three partitioning algorithms on a simple web application using a cloud based NoSQL database management system. The experimental study shows that the techniques presented can improve web application performance.