Study On The Key Technologies Of Geo-Computation On Distributed Clusters

High-performance geo-computation is now a tough problem in Geographic Information System field.Customers use GIS more and more,however they have less tolerance on the responding time of GIS.As a result,the traditional GIS cannot meet the need of customers.Researchers in GIS take the latest and the most popular revolutionary technology in computing into GIS,such as cluster computing,in-memory computing and so on.If we can find a new platform that can proceed geo-computation rapidly in different modes,and can work together with computing platform of the existing GIS,we can improve the performance of GIS.The main work can be determined as below:1.Studied the structure of algorithms scheduling frame in GIS,and concluded the computing and scheduling model.We discussed the target and function of every part of the scheduling frame,and finally realized the frame based on Spark.2.Proposing a new geo-computation algorithm based on the new-designed scheduling frame.We recode a new trajectory sampling algorithm that can run on Spark clusters according to the existing trajectory sampling algorithm.3.Proposing a new scheduling algorithm on Spark.We studied FIFO scheduling algorithm,Fair-Sharing scheduling algorithm and proposed a scale-Fair-sharing scheduling algorithm based on the definition of Fair in Scheduling.And we test these algorithms by an experiment.4.Linked the Spark-based frame to an MPI-based GIS.Designed a GIS structure with Spark-based frame and MPI-based frame parallel processing on it.