190 likes | 338 Vues
JAXA GRID Activity JAXA/EORC Kengo Aizawa KEIO UNIV Hiromichi Fukui. Background. JAXA and KEIO Univ are start to try establishing “Digital Asia” system In Digital Asia, it required to - search and access distributed data - semi-real time data processing
E N D
JAXA GRID Activity JAXA/EORC Kengo Aizawa KEIO UNIV Hiromichi Fukui
Background • JAXA and KEIO Univ are start to try establishing “Digital Asia” system • In Digital Asia, it required to - search and access distributed data - semi-real timedataprocessing JAXA had two of experiments
Purpose • Extract problems to realize Digital Asia system by developing prototype. • Accumulate knowledge, know-how for design of Digital Asia system.
PostgreSQL Assumed scene for Data Grid ・Search satellite data to each node seamlessly ・Databases are geographically distributed ・Each node has different RDB(PostgreSQL、MySQL、Oracle)system USER node(Nepal) Node (Japan) Oracle Node(Chaina) MySQL
PostgreSQL Access to hetero DBMS Oracle My SQL Distributed DB virtual unification Overview DataGrid For Digital Asia Apply (GT3+OGSA-DAI)、access DB over Firewall Unified data view(virtual XML) adopt 「XQuery」 Interface GT3 + OGSA- DAI JDBC FW Unified data view (Virtual XML) Internet SQL実行 XQuery FADGEサーバ SQL SQL実行 XQueryパーザ GT3 + OGSA- DAI GT3 + OGSA- DAI JDBC Query engine results (XML形式) SQL実行 Procesing metadata Metadata for Unified DB GT3 + OGSA- DAI JDBC Translate XQuery to appropriate SQL
System A System B System C Distributed DB virtual unification Overview Data Grid For Digital Asia Create unified data view by Drag and drop XQuery unified dataview (virtual XML) #1 XML XQuery unified dataview #2 Metadata editor for Unified DB XML Columns of each tables will be collected automatically Difference of each DB table colmn will absorbed in the editor XQuery unified dataview #3 XML metadata For unified DB metadata #2 metadata #3 metadata #1 Allows to built virtual unified system for distributed DB
シーン11 シーン1 シーン12 シーン2 Processed image シーン11 シーン11 シーン11 シーン11 シーン1 シーン1 シーン1 シーン1 image processing Image processing シーン12 シーン12 シーン12 シーン12 Image processing image processing シーン2 シーン2 シーン2 シーン2 Computing GRID Overview Management server PC Required spec of server Systemwalker CyberGRIP Master ユーザ ・OS Red Hat Enterprise Linux AS(v. 3)/ES(v. 3) ・Disk 20MB ・memory 30MB + User X 13MB) Processing program sceine Required spec of node ・OS [Windows] Windows SV 2003SE/EE Windows 2000 SV/AS/Pro Windows XP Professional [Linux] Red Hat Enterprise Linux AS(v. 3)/ES(v. 3) Red Hat Linux 9 ・Disk 5MB ・memory 50MB Systemwalker CyberGRIP Agent Systemwalker CyberGRIP Agent Systemwalker CyberGRIP Agent Systemwalker CyberGRIP Agent Computing node PC
Computing GRID Experiment • We estimated following two cases. - submitting 100 dummy* jobs with no data transfer - submitting 100 dummy* jobs with 400MB data transfer * dummy job takes 90sec
Cyber GRIP ・OJC(Organic Job Controller) Allows you to automatically generate a great number of computing jobs with simple descriptions using a script language. It also allows you to do parameter search, parameter sweep, and parametric study. ・GJES(GRID Job Execution System) - Virtually manages dispersed computing resources (computing servers) and equalizes loads on servers. Not only that, if a computer downs during job execution, automatically determines the downed event and executes the relevant job another computer. - Realizing unified job execution in a computing environment with heterogeneous operating systems
Fujitsu plans - To create interface compliant to Globus tool kit
シーン11 シーン11 シーン1 シーン1 シーン12 シーン12 シーン2 シーン2 Future Plan Search Analize Computing Grid Data Grid モザイク処理 User シーン11 シーン1 モザイク処理 シーン1 変化抽出 シーン1 シーン12 シーン2 モザイク処理 Grid Portal Server 変化抽出 モザイク処理 Computing GRID Data GRID (検索) 各受信局 Realize GRID from search to Analize