Accelerating geospatial database services with Graphical Processing Units

Abstract

With the growing need of instant or almost instant processing and retrieval when working on large data-sets we ask ourselves the following question “What impact would switching from a conventional CPU database to a GPU accelerated database have on emergency systems using large geospatial data-sets”. [Methodology] We chose to use Design Science and more specifically the method called optimization. [Motivation for the study] We observed a knowledge gap in the field of geospatial analysis regarding use cases associated with emergency systems and with new technological advances both in software and hardware there is a need to reevaluate current systems. [Test results] The result displays that GPU accelerated databases and SPARK databases do not increase the efficiency of processing and retrieving of large geospatial data. [Discussion] Even if we expected the GPU accelerated database to perform better than the standard CPU database we could not see any benefit from switching to a GPU accelerated database or SPARK database. [Conclusion] Our tool we created did enable us to take more informed decisions when making decisions on what database is best for our use case, we did, however, conclude that there is no benefit for us to switch to a SPARK or GPU accelerated database. [Future work] We found several things that would benefit further research in our area, Both in technology and scale.