Hardware Capacity Evaluation in Shared-Nothing Data Warehouses

Parallel data warehouses have mostly been seen as dedicated decision support systems, but as the need for cost effective multi-application business solutions grows, non-dedication and adaptation to existing environments become evermore tempting. Data placement in such systems is a major problem seeing as existing environments may have heterogeneous nodes, which process data at different rates, and non-dedicated environments can not promise full resource commitment to execute a query. Therefore it is critical to distribute a data warehouse´s information in such a way that no node be overloaded or underestimated. This paper presents the capacity evaluator (CE), an application capable of measuring the system´s ability to process information, which in turn helps the data warehouse parallel architecture´s (DWPA) automatic data placer determine exactly how much data should be allocated to each node.