Performance comparison of GES DISC data as a service between server-based system and cloud system

The NASA Goddard Earth Science Data and Information Service Center (GES DISC), in cooperation with the Goddard Information Technology & Communications Directorate, demonstrates and evaluates provision of "Data-as-a-Service" in a cloud environment using the OPeNDAP (Open-source Project for a Network Data Access Protocol) protocols. The demonstration requires porting the OPeNDAP software to the cloud platform along with a representative set of data and then exercising the server using several clients. The evaluation examines two aspects of using open source software in the cloud to serve large volumes of satellite data for public access and simple subsetting: a) Ease of porting and operating OPeNDAP in the Goddard Cloud and Amazon Elastic Cloud Computing (EC2) and Simple Storage Service (S3) environments, including evaluation of the time needed to setup one instance; b) Access performance, e.g. data access stability and speed of the cloud environments as compared to existing GES DISC capabilities. Four kinds of satellite data products with different data formats (HDF4, HDF5) were selected as the test data: Advanced Infrared Sounder (AIRS) on the Aqua satellite, Ozone Monitoring Instrument (OMI) on the Aura satellite, Tropical Rainfall Measuring Mission (TRMM), and Modern-Era Retrospective Analysis for Research and Applications (MERRA). For each product, 25 granules were used to test access stability and speed. The Giovanni (GES-DISC Interactive Online Visualization ANd aNalysis Infrastructure) and GrADS (Grid and Analysis System) data services were also deployed to the cloud platforms to compare the data analysis performance between existing systems and cloud systems. We also evaluated the challenges to migrating these services to the cloud architectures examined.