Archived: 30 January 2006
University of Adelaide Masters by coursework thesis, November 2004.
Supervisors: Andrew Wendelborn and Paul Coddington
Resource Brokers on the Grid consult a number of distributed information services to select the best data source and/or computational resource based on user requirements. The consultation task increases the design complexity of Resource Brokers.
The Distributed Grid Query Service (distributedGQS) project is the first attempt to unify distributed grid information services that any Resource Broker needs on the Grid. The services are: Metadata services which provide unique logical file name out of metadata descriptions; Data Replication services which provide storage identifiers, possibly replicated, out of unique logical file name; Information services which provide resource information out of user requirements; and Network Information services which provide the network status and performance to a certain host.
In this current implementation, Metadata services and Data Replication services are being encapsulated within the distributedGQS service. These services are layered on OGSA-DAI Grid services, which are generic data access and integration services. Open Grid Services Architecture - Data Access and Integration (OGSA-DAI) Grid services provide uniform interfaces to heterogeneous databases. The modification of the information data sources and the security of the distributedGQS service are governed by using a robust and reliable software, the Enterprise Level Data Access Services (ELDAS) project.
Grid Resource Brokers need to consult one service, the distributedGQS service, to obtain information about grid resources. Data that matches user requirements are retrieved using newly developed generic APIs that abstract over the underlying middleware. A Web-based user interface to the unified services has also been developed, the distributedGQS user portal.
Within the framework of this project, distributed metadata services has been investigated, namely the Metadata Location Service (MLS) which consists of Metadata Catalog Index (MCI) that indexes Local Metadata Catalog (LMC); MCI can be used as a stand-alone component indexing existing Metadata Catalogs. Moreover, an initial Metadata Schema, describing metadata (data about data) attributes, has also been developed for Belle experiment; a high energy physics experiment. Furthermore, to visualize the interacting components, and trace the design and data communications of distributed services, Simple Promela INterpreter (SPIN) distributed systems model checker, is used to visualize and trace the data communications of the encapsulated Data Replication Services; Replica Location Service (RLS), via channels and synchronous communication mode.
The distributedGQS service decreases the design complexity of Resource Brokers, and improves the overall performance of Grid-based applications.