The Retrieval And Interpolation Of ARGO Ocean Environment Observation Data Based On Cloud Platform

With the development of cloud platform, big data has attracted more and moreattention. Because the analysis of the real-time big data needs a framework such asMapReduce which assigns jobs to hundred, thousand or even thousands of computers.So the analysis of big data is always closely combined with cloud platform.Technically speaking, the relation between big data and cloud computing can bedescribed as inseparable as positive and negative sides of the same coin. Obviously,big data cannot be processed by a single computer, so big data must adopt adistributed computing architecture. This architecture is mining massive data, but therealization of it must rely on cloud computing such as distributed databases,distributed processing, cloud storage or virtualization technology.With the development of cloud computing, open source cloud platform Hadoopapparently has become one of the world’s most important clouds computingtechnology, Hadoop MapReduce in parallel computing model is also used indistributed computing. The framework separates massive data and assigned them tomultiple nodes, and then parallel computes each node, and merges and outputs theresults. The output of the stage is the input of the input. Therefore, you can images adistributed computing structure similar to a tree structure. In each of the differentstages,it will produce different outputs,in the same time,in a distributed clutercomputing resources the serial and combination can also be well processed.ARGO plays a very important role on climate protection and the prevention ofmarine disasters. ARGO data belongs to typical big data, but the software related toARGO data is rarely. Because ARGO data have characteristics such as a amount ofdata, discrete distribution. In order to comparing with distributed patterns of otherspace phenomena, the paper mainly make spatial interpolation for ARGO data, inwhich ARGO discrete point data can be converted to a continuous discrete pointdata surface. The different interpolation methods will produce different efficiency, so this paper realizes two methods of interpolation.One way is that through the MapReduce distributed computing model ARGOdata is separated into several blocks,and then each block of ARGO data isinterpolation computed, in which derives a continuous data interpolated surface. Thesecond method is carried out on the basis of the first method. In this method, we buildan index table for each data of each block by KD tree, thereby reducing the number ofsearching points, increasing the efficiency of the interpolation.In this paper, the experiment collected a lot of ARGO data in order to comparingthe computing efficiency of directly interpolation algorithm with the efficiency ofKD-tree index interpolation algorithm. The results also show that KD-tree indexinterpolation algorithm has a higher efficiency than the directly interpolationalgorithm.The method has achieved good results on the ARGO data analysis.