OGSA-DAI User Group Infrastructure Breakout 1 Objective – To Propose… • The collection of WS-* specifications to base OGSA-DAI on • The level of support that OGSA-DAI should provide for the evolving GGF DAIS specification • The deployment platforms that OGSA-DAI should support 2 Process • We can’t do everything in OGSA-DAI due to finite time and budget so we must prioritise – Requirements for projects – Requirement across projects – Requirements of high priority • Today we should try and avoid the politics and keep and eye on the clock 3 WS-* • Generally – OGSI – WS-RF – WS-I • Which need to support – – – – – – – Identity of resources Stateful interactions Meta data Sessions Lifetime management Endpoint fragility Security • Visibility of the differences – What is the cost (to the administrator/consumer) of different approaches 4 • Cost of maintaining incompatible infrastructure higher than writing access code • WS-I OGSA-DAI followed by WS-RF OGSA-DAI • WS-I/WS-RF can be used to represent the same data resources in different ways • Dangerous to track evolving specs • People want GT4 compatible implementation – But to require GT4 could also be problematic • Can OGSA-DAI produce a straw man for the way forward for WS-I/WS-RF 5 • Identifying data resources – Applications have different ways of naming data and providing meta data for data resources – Not a clear separation between data and meta-data. eDiamond access mechanisms are the same – Be useful to be able to associate arbitrary meta data with data resource – Mapping from name to service location • Life sciences have LSIds • Need a hierarchy of names/mappings will help map down to a data resource. (GSH vs GSRs ?) • What are we mapping to. Resource, Services, datasets. • Major performance hit if you have to map name before doing anything • Gnerally rely on registry to map from name to resource/service 6 • Security – OGSI/WSRF say nothing about security – GT3 imposed GSI security model 7 Support For DAIS • OGSA-DAI current – Perform document model – GDSF = data resource – GDS = session over data resource • DAIS – – – – 8 Data service = data resource Direct access via RPC style messages Indirect access via derived data services Other specs include • DAIS – Be useful to hide the model switch to DAIS behind client toolkit – Staged approach • Simple WS-I • More capable WS-RF 9 Deployment Platform • Build data services within – – – Globus Toolkit GT2/3/4 Unicore J2EE • – • Windows Linux Other UNIX flavour Point data services at – – – – – – 10 Java C++ C# Perl Deploy data services on – – – • .Net Build data services with – – – – • Axis/Tomcat DB2 Oracle SQLServer Postgres MySQL Xindice • Does using WS-I imply not using Globus? – Not necessarily but maybe we assume Tomcat/Axis • Three axes of globus – What you build on – What clients require – What is promised to those you call • Client library may need porting to different languages • Some users won’t use client library, e.g. resources included in BPEL. – MyGrid/Gold want to ensure that DAI rendered resources can appear in workflows • • • • May want to call Perl on the server side Files Object DBs – no one here doing it OGSA-DAI in front of continuous data sources – Originally envisaged but no one doing it with ogsa-dai yet 11 Summary Only 9 people in the room 4 of whom were users from 5 projects 12 Infrastructure • The community is uncertain about what to do • Provide straw men to show implications • Some people want WS-I others GT4 (!) others don’t want to have to care • WS-I first, WS-RF later as it matures 13 DAIS • Keep functions that have provided value – E.g. extensibility of Perform Documents, i.e. activities • Validate the DAIS standard – Not a high priority user req • Composability with other WS specs 14 Deployment • Files Files Files – Some subclass, e.g. CSV, DFDL – When the file is the table • Perl • What do we mean by WS-R/WS-I? – Standards or Platforms • Different client library languages – Java, C++, C#, Perl, C?, Fortran? – As a contribution • Publish an API for the client library 15