As cloud based web applications become more common, developers are increasingly turning to cloud based NoSQL database solutions. While optimizing traditional relational database management systems is well studied and documented, there is in comparison little written about optimizing cloud based NoSQL databases. We present a system that partitions a cloud based database using data cluster analysis on database transaction log files. The intent is to move the partitions to regional edge servers, thus reducing the physical distance from the client to the data. We were successful in significantly reducing the average response time of page requests compared to that of a cloud based DBMS alone or round robin partitioning . Our system provides an advantage over traditional partitioning methods because it is able to dynamically find regional trends based on the types of queries that originate from a particular region.
展开▼