T.-K. Kim, S.-H. Jung, K.-R. Kim, J.-S. Yoo, and W.-S. Cho (Korea)
Database Cluster, Data Warehouse, OLAP, Query Processing
As the overall amount of data increases rapidly in most areas of business, performance of complex OLAP queries used in significant decision-making support has been in decline. To overcome this challenge, most companies invest more in their hardware infrastructure. We propose a high performance database cluster system called HyperDB to solve this performance issue. HyperDB is a virtual database system running on top of internet connected PCs; the PCs are used for their own purposes at ordinary times, but they are able to participate in a database cluster system during non-office hours. Note that conventional database cluster systems use a static number of tightly-coupled dedicated nodes, and therefore there is almost no such flexibility and extensibility. We propose a logical full replication technique and an inter-intra mixed query routing technique for extensibility and performance. Note that both techniques are known as difficult problems within the conventional database cluster systems. Experiments on the TPC-R benchmark shows significant performance increases compared with conventional approaches.
Important Links:
Go Back
