Vol.2 No.1 2009

43/88

Research paper : How Grid enables E-Science? (Y. Tanaka)−40−Synthesiology - English edition Vol.2 No.1 (2009) file system used is Gfarm v1.4. At this point, all ASTER data (about 140 TB) is stored, and about 70~100 GB of data are transmitted from NASA daily to be stored in the GEO Grid cluster. For management of metadata, PostGIS, expanded for GIS for PostgreSQL, is used, and the metadata is provided as a data service by OGSA-DAI. For data processing, an F32 cluster composed of 256 nodes of dual Xeon processors connected by Giga-bit Ethernet is used. The GEO Grid cluster and the F32 cluster are connected by a 10 Giga-bit Ethernet.A demonstration experiment was conducted for the GEO Grid that integrated multiple databases managed by multiple organizations, by constructing a “DB Collaboration Application” that enabled simultaneous search of AIST’s ASTER data, MODIS (Moderate Resolution Imaging Spectroradiometer)[22] data, and Formsat2 data owned by the National Space Organization (NSPO) of Taiwan. It was confirmed that “data search,” “conversion to digital geographic model,” and “transfer of result,” could be implemented as individual services. These were provided as high level services by linkages based on authorization provided by GSI and VOMS, a VO was constructed through integration, and the research environment was then provided to the application community. Also, through applications that allow linkage of ASTER data and free sensor data, and an application that returns the search results in WFS, it was confirmed that the proposed security architecture was able to provide appropriate access control based on the requirements of application and service providers.6 DiscussionHere, I will discuss the findings obtained through the construction of the system and its subsequent operation.Through the preliminary assessment of each software component and the development and testing of the ASTER Grid system, we were able to verify the adequacy of both the design and the implementation of the GEO Grid information infrastructure. Using the security framework utilizing GAMA and VOMS, it was confirmed that the system provided an easy-to-use interface to users, and flexible access control in compliance with the policies of the service providers. Also, through access control at the VO level using VOMS, a security system scalable to the number of users could be achieved. By using existing software and tools such as GAMA, VOMS, OGSA-DAI, and the Globus Toolkit, we were able to construct the system with low development costs. Other than the portlet used for applications, the development done by the information infrastructure side was an incorporation of VOMS and GAMA interfaces into the GridSphere. By designing and implementing all basic technologies based on a standard protocol and a standard interface, we were able to easily construct a high-grade system by linking multiple, independent basic technologies.For the construction of an actual grid system, this may require development and use of giant middleware packages such as the data analysis system for the large-scale accelerator experiment in high-energy physics using the gLite Grid Middleware[23] suite developed by EGEE of Europe, or the Cyber Science Infrastructure project where a large-scale research grid has been constructed over university and research institutes using the NAREGI middleware[24] suite developed by the National Research Grid Initiative of Japan. The Earth System Grid[25] and GEON[26] of the USA also intend to construct a research environment with integrated earth observation data as in the GEO Grid. All of these use authorization and approval based on grid security in some part, but most are composed of non-grid technology based on web services. A case like GEO Grid is highly original, where the research environment is constructed with flexible access control based on grid security by creating a VO through the combination of data and computation provided as a service using various grid middleware suites. Therefore, this research is significant because it demonstrated that a large-scale system could be readily built by linking the grid middleware, as long as all elements were implemented according to standard security and protocols, as described in this paper.Through building this system, it has been confirmed that the basic technologies can be utilized without any significant problems. As issues that must be solved toward full realization, the following five can be pointed out.(1) Creation of a tool kitMany of the grid middleware suites used had complex installation and configuration requirements, and cannot be readily installed and used by everyone. Although the user is provided with a simple interface that can be used with only a user name and password, it is necessary to provide a tool kit where the required middleware set can be easily designed and installed for the benefit of all participants, including service providers and VO administrators, in order to promote further development in various application fields.(2) Realization of a more flexible authorization functionSome of the existing application communities already employ their own authorization function, such as OpenID currently used in some bio information fields. To achieve seamless transfer from the existing research environment to an E-Science environment, it is necessary to implement a more flexible authorization function that produces grid authorization information from the authorization mechanism that is already being used, such as those of OpenID, Shibboleth, or Kerberos.(3) Construction of workflowsMany of the application researchers use certain preset

※このページを正しく表示するにはFlashPlayer9以上が必要です