<![CDATA[Data partition and the accordingly index technologies which could result in uniform data distribution and fast data finding are critical in high parallelism for shared nothing architecture to minimize the transaction processing time. An improved-hash based multi dimensional index mechanism is present in this paper to achieve high parallelism performance for distributed data-parallel computation in shared nothing architecture. After partitioning and storing data using improved-hash function based on partitioning column, the multi dimensional indexes based on multi columns and the corresponding data lookup procedure are constructed. Afterwards, the space complexity and time complexity are analyzed. DOI: http://dx.doi.org/10.11591/telkomnika.v11i4.2359 ]]>
Copyrights © 2013
