The Developer Perspective GEMEDA Grid-Enabled Microeconometric Data Analysis Design • initial plan was to extend an existing demonstrator that had grid-enabled a related problem • start point: review of current technologies • ad-hoc requirements capture: dialogue between developers and users – regular meeting of the full project team – one-to-one meetings/communication for specific issues : data, modelling, presentation Technical Strategy • Hitch onto established trends & ongoing investments: Web service technologies (portals avoid local deployment of complex middleware stacks) OGSA-DAI NGS Oracle for databases, MPI for parallelism The establishment of a critical mass of compatible Grid-enabled tools is a necessary condition for success of e-Social Science in the UK Development Issues • small scale project with funding, time and scheduling constraints • e-Science research staff recruitment • reliability, immaturity of documentation, and error reporting for middleware services (esp. NGS) • software re-use (human capital) – Athens authentication code from ConvertGrid – GeoTools (java mapping library) from Hydra I Grid • dialogue between developer and user (e.g. what is a variable?) Lessons Learnt • development is hostage to staffing issues • what the user(s) want isn't necessarily what they need • specialist elements of the service are often the easiest (in a relative sense) to develop – parallel code – visualization displays. • linking everything together is the difficult bit Software Used: Linux Mandkrake 10.1 Java 1.4.2_06 Tomcat 5.0.28 Globus Toolkit 3.2 OGSA-DAI 4 Java Cog Kit 4 Spring J2EE application framework 1.2.3 DWR Ajax library ProFTP Oracle 9i RAC edition Python 2.3 (dataset creation & wrapper to second Java instance & Cog Kit code) Eclipe IDE Fortran95+MPI C (utiliies for converting output for use with the map) GeoTools (Java library for mapping) Future Plans • The project has now concluded – Ensuring long term sustainability is a problem – The plan is for the service/portal to become part of the NCeSS demonstrator suite. – Re-engineering may be required. (e.g. for Globus WSRF compliance) NGS Oracle hosted datasets NGS HPC computing nodes Manchester, Oxford, Leeds, CCLRC RAL BHPS GEMEDA parallelized SARs SARs Household Individual analysis code grid service NGS MyProxy server OGSA-DAI Grid Security Infrastructure DAISGR Globus Toolkit Core (DAI Service Group Registry) GEMEDA Architecture Diagram Axis Athens Server Apache Tomcat GEMEDA FTP server Apache Tomcat Spring Framework JSP Web interface + Results visualization applet