Issue No.04 - July/August (2009 vol.11)
James L. Johnson , Western Washington University
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MCSE.2009.127
In a cloud computing context, the MapReduce algorithm comprises two massively parallel operations linked by a generic sorting and data-distribution process. Although this algorithm is the workhorse in most cloud computing strategies, it's a special case of a more general dataflow. In place of the two cloud operations, the proposed method substitutes longer sequences and then lets the user direct outputs to any subsequent downstream operation. However, the method retains the job-supervisor infrastructure, which performs the necessary sorting, collating, and distributing of these outputs prior to initiating operations. To evaluate SQL database queries, particularly those with correlated subqueries, a computation identifies and aligns data elements from widely separated storage locations, suggesting cloud algorithms that exploit the supervisory sorting process to achieve the desired alignments. Exploring such algorithms reveals that a few customizable templates, assembled recursively as necessary, can handle a wide class of SQL data-mining queries.
cloud computing, distributed computing, dataflow architectures, database structures, database query evaluation, SQL cloud algorithms
James L. Johnson, "SQL in the Clouds", Computing in Science & Engineering, vol.11, no. 4, pp. 12-28, July/August 2009, doi:10.1109/MCSE.2009.127