Parallelization of web processing services on cloud computing: A case study of ...

• Autor(es): Murillo, Carlos Andrés Osorio
• Data: 2011
• Identificador Persistente: http://hdl.handle.net/10362/8294
• Origem: Repositório Institucional da UNL
• Assunto(s): Web Processing Services; Parallelization Algorithms; Interpolation; Geostatistics; Cloud Computing
• Tipo de Documento: Dissertação de Mestrado
• Idioma: Inglês
• Orientador(es): Guijarro, Joaquín Huerta; Remke, Albert; Painho, Marco

Descrição

Dissertation submitted in partial fulfillment of the requirements for the Degree of Master of Science in Geospatial Technologies. In the last decade the publication of geographic information has increased in Internet, especially with the emergence of new technologies to share information. This information requires the use of technologies of geoprocessing online that use new platforms such as Cloud Computing. This thesis work evaluates the parallelization of geoprocesses on the Cloud platform Amazon Web Service (AWS), through OGC Web Processing Services (WPS) using the 52North WPS framework. This evaluation is performed using a new implementation of a Geostatistical library in Java with parallelization capabilities. The geoprocessing is tested by incrementing the number of micro instances on the Cloud through GridGain technology. The Geostatistical library obtains similar interpolated values compared with the software ArcGIS. In the Inverse Distance Weight (IDW) and Radial Basis Functions (RBF) methods were not found differences. In the Ordinary and Universal Kriging methods differences have been found of 0.01% regarding the Root Mean Square (RMS) error.The parallelization process demonstrates that the duration of the interpolation decreases when the number of nodes increases. The duration behavior depends on the size of input dataset and the number of pixels to be interpolated. The maximum reduction in time was found with the largest configuration used in the research (1.000.000 of pixels and a dataset of 10.000 points). The execution time decreased in 83% working with 10 nodes in the Ordinary Kriging and IDW methods. However, the differences in duration working with 5 nodes and 10 nodes were not statistically significant. The reductions with 5 nodes were 72% and 71% in the Ordinary Kriging and IDW methods respectively. Finally, the experiments show that the geoprocessing on Cloud Computing is feasible using the WPS interface. The performance of the geostatistical methods deployed through the WPS services can improve by the parallelization technique. This thesis proves that the parallelization on the Cloud is viable using a Grid configuration. The evaluation also showed that parallelization of geoprocesses on the Cloud for academic purposes is inexpensive using Amazon AWS platform.