This documentation refers to the Iteration 5 release of the AstroGrid Grid Data Warehouse.

The AstroGrid Grid Data Warehouse (GDW) is intended to be a facility for storing and (cross-)querying both standard astronomy datasets and user-supplied datasets.

At the present time, it provides read-only access to pre-stored datasets; user-supplied dataset functionality will be forthcoming.

The GDW uses the OGSA-DAI grid middleware for database access and querying. This means it also currently relies on the OGSA implementation supplied with the Globus Toolkit 3.0. Both of these technologies must be deployed using Apache Tomcat.

Important note for early adopters

The Iteration 5 release of the AstroGrid Grid Data Warehouse is hardwired to expect an underlying Postgres 7.3 DBMS, where all the datasets accessed via the GDW contain columns named ra (right ascension) and decl (declination).

There is also a hardwired limit of 100,000 on the number of rows returned by any query.

The next release of the GDW should support user-customisable ADQL-SQL translation (not just the current Postgres-flavoured translation), with user-configurable column names for right ascension and declination, and user-configurable limits.

At present, the Astrogrid Grid Data Warehouse (GDW) is a specialised deployment of the standard Astrogrid DataCenter. All required GDW front-end components are distributed as part of the AstroGrid DataCenter package, distributed as PAL-{version}.war. After installation, this package will need certain minor reconfigurations to run as a GDW (see below).

To supply its actual database access, the GDW requires an installation of the OGSA-DAI grid middleware. The Iteration 4.1 release of GDW assumed OGSA-DAI 3.0; the Iteration 5 release assumes OGSA-DAI 3.1 (the latest release at the time of writing).

OGSA-DAI itself currently relies on the OGSA implementation (v. 3.0.2) in the Globus Toolkit 3 core distribution from the Globus Alliance.

Both of the above packages come with comprehensive installation instructions; a vanilla installation of both should work with the GDW.

Important note: At present, it is required that the OGSA-DAI (and thus OGSA) installation be co-located on the same host as the GDW installation. (They do not have to be running in the same tomcat instance, however). This is because local file storage is currently used for delivery of results from OGSA-DAI. Future releases of the GDW will not have this limitation.

To install a GDW, first download and install a standard AstroGrid DataCenter component; the DataCenter component provides comprehensive installation and post-install test instructions in its PAL install documentation.

To configure the installed datacenter as a GDW, the following entries need to be added or adjusted in the tomcat configuration file \$\{CATALINA_HOME\}/conf/datacenter.properties, where \$\{CATALINA_HOME\} is the tomcat toplevel directory for the tomcat in which the DataCenter webapp is installed:

# Configure DataCenter to use GDW plugin
# (DELETE ANY OTHER ENTRIES FOR datacenter.querier.plugin)
datacenter.querier.plugin=org.astrogrid.datacenter.queriers.ogsadai.WarehouseQuerier 

# Location of OGSA-DAI grid service registry
# CONFIGURE THIS TO MATCH YOUR OGSA-DAI INSTALLATION
WAREHOUSE_OgsaDaiHostString=http://hydra.star.le.ac.uk:8082
WAREHOUSE_OgsaDaiRegistryString=/gdw/services/ogsadai/DAIServiceGroupRegistry

# Location of XSLT stylesheet for XML RowSet -> VOTable transformation
# This is a one-size-fits-all hack that will need replacing.
# NO NEED TO CHANGE THIS
WAREHOUSE_XslTransform=http://astrogrid.ast.cam.ac.uk/xslt/ag-warehouse-first.xsl

# Location of JVM for shelling out to GdsQueryDelegate
# CONFIGURE THIS TO MATCH YOUR JAVA INSTALLATION
WAREHOUSE_WarehouseJvm=/usr/local/java/j2sdk1.4.1_02/bin/java

# Location of executable jar for shelled-out ogsadai call
# CONFIGURE THIS TO MATCH THE WEB-INF/classes DIRECTORY IN YOUR OWN
# DATACENTER INSTALLATION
WAREHOUSE_ExecutableJarPath=/home/data/gdw/tomcat1/webapps/pal-SNAPSHOT/WEB-INF/classes/

# CHANGE {VERSION} TO MATCH THE NAME OF THE EXECUTABLE JAR FOUND 
# IN THE WAREHOUSE_ExecutableJarPath DIRECTORY OF YOUR DATACENTER INST.
WAREHOUSE_ExecutableJarName=astrogrid-ogsadai-app-{VERSION}.jar