Transcript ppt
SRB + Web Services = Datagrid Management System (DGMS) Arcot Rajasekar (Arun Jagatheesan) San Diego Supercomputer Center National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center Outline SRB WSDL-based Services: • What do we plan to accomplish? • Why are we doing this? • How do we plan to reach our goals? • Design, Arch, Implementation • Who is doing this? • When? • What are our timelines National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center What? • SRB = Storage Resource Broker + MCAT • Transparent access to distributed storage and services • Used and proven technology • • • • • • • More than 3 years of deployment over 2 dozen projects Mol Sci, Neuro Sci, Astronomy, Med, ESS, DigLibs, … ~18 TB of data handled until last year >7 million files Multiple Access Methods – APIs, GUIs, Commands, Web-HTTP Developed at SDSC/NPACI National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center Why • SRB WSDL Services Universal DataGrid Services • Emergence of Service oriented architectures • To be compliant with the Grid Requirements • SRB as a Grid Component • SRB as a Commodity service • Data Management as a service in the Grid • Project-driven • TeraGrid, NVO, PPDG, GriPhyN, SCEC, … • BIRN, RoadNet, IT2, ... • Loose coupling, interoperability, standards based. National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center 1 Min Tutorial • Web Page (HTML) • Web Service • Searched and used by human being • Any computer • HTML • HTTP • Google • Searched and used by computer programs • Any programming language, OS etc • XML/ WSDL – Web •Service WSDL Description • SOAP (HTTP/SMTP) • –SOAP Transport/Acces (HTTP/SMTP) • UDDI - Discover • UDDI National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center How - Design • SRB = Command Channel + Data Channel • Protocol = set of rules for end points to communicate with each other • SRB Web services – Protocol for exchange of SRB Command Channel ONLY. • Client flexibility to choose any protocol (SRBRPC, FTP, GFTP, HTTP) for local data channel • Internal Data grid operations like replication – always SRB-RPC National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center How – Design II • All or nothing (Atomic) • Value addition - Services apart from original SRB Services • Asynchronous – UUID based sessions • Session and event management • WSCAT – (We ♥ databases?), Status Query Service • Security – CA, HTTPS, GSI-SOAP etc., • Open protocol for SRB Web services client. • Each operation a service – not a single aggregated SRB Web service. National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center How - Architecture SOAP Server SRB service bean Java Native Interface JDBC Query Interface and Trigger Mechanism Service Publisher and WSDL documents SRB Native Methods Transaction or Event Database Current Implementation. Will change soon. SRB Server Jasmine GridFTP SRM Any DBMS or Data Manager WSDL- DataGrid National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center Issues + Status • SOAP client interoperability • Asynchronous call – Management (UUID??) • Usage of www.schemas.UniversalDatagrid.org? • Some preliminary version of WSDL • Ingest File to Grid, srbOperations. • Core Datagrid Operations Document in creation • Schemas for Datagrids National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center Who • SDSC DAKS/NPACI DICE • SDSC GridPortal Group • Related Work: • Jlab • LBNL • … • We plan to collaborate to get a common service definitions National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center When • First Version (GGF5): Data Movement Functions Put, Get and Replicate • Second Version (Jan 2003): Data Discovery • Query, Browse, Attach Metadata, Extract Metadata,… • Other Versions (2003-2004): More SRB functions exposed through WSDL framework • After That : Layers National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center What lies ahead? Layers Knowledge Management & Information Mediation Services Data Mining Services Data Management Services National Partnership for Advanced Computational Infrastructure San Diego Supercomputer Center