8:50 - 10:00 Keynote

Welcome

Gordon Bell

10:30 - 12:00 Grid Applications and Data Grid

Experiences in Developing a Node of an International Computational Physics Data Grid, Paul Coddington, Gerson Galang, Waseem Kamleh, Derek Leinweber, Sam Moskwa, Julia Patterson, Qiang Wang, Andrew Wendelborn, Shunde Zhang and Qunfang Zhang

Grid services for e-archaeology, Oystein Pettersen, Nicole Bordes, Sean Ulm, David Gwynne, Terry Simmich and Bernard Pailthorpe

Commodity-Grid Based Distributed Pattern Recognition Framework, Anang Hudaya Muhamad Amin and Asad I. Khan

1:00 - 3:00 Middleware and Web Services

State Aware WSDL, Michael Brock and Andrzej Goscinski

Service Migration in Autonomic Service Oriented Grids, Michael Messig and Andrzej Goscinski

Supporting Large Scale eResearch Infrastructures with Adapted Live Streaming Capabilities, Syed Hasan, Laurent Lefevre, Zhiyi Huang and Paul Werstein

Invited talk on ARCS and NCRIS

3:30 - 4:30 Grid Resources and Performance

Resource Evaluation and Node Monitoring in Service Oriented Ad-hoc Grids, Ian Scriven, Andrew Lewis, Matthew Smith and Thomas Friese

Grid Resource Allocation: Allocation Mechanisms and Utilisation Patterns, Stefan Krawczyk and Kris Bubendorfer

4:30 - 5:30 Group discussion

7:30 - 11:00 ACSW conference dinner

Lagoon seafood restaurant, Northbeach

Keynote

Gordon Bell is a principal researcher with the Microsoft Research Group, San Francisco (1995-) working on a system, MyLifeBits to capture everything in a person’s life. His career includes: vice president of R & D, Digital Equipment Corp. (1960-1983); Professor of Computer Science and electrical engineering, Carnegie-Mellon University (1966-72); founding Assistant Director of the National Science Foundation's Computing and Information Sciences and Engineering (CISE) Directorate (1986-1988); National Research and Education Network (NREN) panel chair (1987-1988) for creating the internet; advisor/investor in 100+ start-up companies. Since 1987 he has sponsored the ACM’s Gordon Bell Prizes for parallelism awarded annually at Supercomputing. He has BS and MS degrees from MIT (1956-57), University of New South Wales Fulbright Scholar (1957-58) and is a member of the AAAS, ACM, IEEE, and the National Academies of Engineering (1977) and Sciences (2007). Awards include: ACM-IEEE Eckert-Mauchly Award, the IEEE’s Computer Pioneer and IEEE McDowell Awards, and the IEEE Von Neumann Medal (1992),and The 1991 National Medal of Technology "for his continuing intellectual and industrial achievements in the field of computer design; and for his leading role in establishing ... computers that serve as a significant tool for engineering, science, and industry."

Abstracts

Experiences in Developing a Node of an International Computational Physics Data Grid, Paul Coddington, Gerson Galang, Waseem Kamleh, Derek Leinweber, Sam Moskwa1, Julia Patterson1, Qiang Wang, Andrew Wendelborn, Shunde Zhang, Qunfang Zhang

The International Lattice Data Grid (ILDG) is an international collaboration that creates standards to enable sharing of data produced by Lattice Quantum ChromoDynamics (QCD) simulations, which are very computationally expensive. In this paper we summarize our experiences in developing an Australian node of the ILDG, including implementing a program to generate metadata conforming to the QCDml XML schema developed by the ILDG, storing the metadata in an XML database, developing metadata catalog and file catalog services using standard web service interfaces, providing data download using an SRM interface, and implementing a web portal to the ILDG. One of the problems we encountered in this work was that existing metadata catalog software was not able to meet the requirements of this project. Although this work is specific to Lattice QCD data and the ILDG specifications, most of the main issues in developing the data grid node are quite general and not specific to the type of data.

Grid services for e-archaeology, Oystein Pettersen, Nicole Bordes, Sean Ulm, David Gwynne, Terry Simmich and Bernard Pailthorpe

Archaeological data collection is based on the description of archaeological contexts. An archaeological excavation demolishes the original matrix within which the cultural material is found and special care is taken to record spatial context. Each artifact is described in terms of its physical and spatial properties as well as its relation to the matrix (for example soil composition). As several thousands of artifacts can be unearthed during a field season, there is a need to develop digital resources and collections that focus on the publication and preservation of data and the creation of tools for the analysis of these data. The first section of this paper presents preliminary results and the lessons learnt on the development of a prototype for an Australian archaeological digital collection based on data grid middleware and infrastructure. The second section introduces a versatile 3D reconstruction tool that visualizes the excavated archaeological artifacts with its associated stratigraphy. The data come from two major archaeological projects in Queensland, Australia: the Mill Point Archaeological Project and the Cania Gorge Regional Archaeological Project. These case studies were selected to represent the different challenges in deploying these digital technologies to Australian archaeological applications.

Commodity-Grid Based Distributed Pattern Recognition Framework, Anang Hudaya Muhamad Amin and Asad I. Khan

Large-scale pattern recognition for data mining requires significant processing resources. Distributed pattern recognition provides an avenue for achieving large-scale pattern recognition by using a state-of-the art data classifier for fast tracking large-scale data analyses. In this paper, we will introduce a framework for distributed pattern recognition which is grid enabled and employs a distributed single-cycle learning Associative Memory approach. The framework comprises commodity-grid network for pattern recognition processing using the single-cycle approach. Our research has shown that the distributed pattern recognition using this framework will provide a fast and reliable resource for use in data mining. Our work also shows that the commodity-grid provide an easy-touse front-end for accessing a distributed system supporting complex operations.

State Aware WSDL, Michael Brock and Andrzej Goscinski

With the recent innovations in stateful web services, they are now being used to support the construction of distributed systems using software as a service. While the state of web services is preserved, the state is still hidden from clients thus searches for both functionality and state remains a two step process. Proposed in this report is the Resources Via Web Instances (RVWI) framework. RVWI grants to web services the ability to include their state and characteristics in their WSDL. This was done by allowing snapshots (instances) of a web service to be listed in the WSDL of the web service. Instances were utilised as they contain state and characteristic information directly from the web service. Thanks to the inclusion of state and characteristics, queries for web services can now be carried out on the availability of a web service and the ‘dimensions’ of resources.

Service Migration in Autonomic Service Oriented Grids, Michael Messig and Andrzej Goscinski

The introduction of Web services into grids helped to address their two main obstacles to be embraced by business and industry, heterogeneity and useability. However, many problems are still open, e.g., grid reconfiguration, reliability and computing optimization. We argue here that a mechanism that could help solving these problems is Web Service migration, a part of automatic and transparent brokerage. Web service migration presents a number of new requirements not addressed in traditional process migration, being the outcome of Web services specific configuration of hosts and application servers, and availability of Web services / resources state. In this paper we report on our study into the development of a Web service migration facility focused on providing migration of services in a Service Oriented Grid environment. We present a novel approach to Web service migration, embodied in a System Management Broker, which is transparent, interoperable and flexible. We take the requirements of Web services into consideration when discovering suitable destination hosts and match services to suitable grid resources which are able to fulfil the needs of the service. We discuss a number of experiments conducted with different types of grid and Web service applications to highlight the feasibility and effectiveness of our migration facility and demonstrate how our facility significantly improves Service Oriented Grids.

Supporting Large Scale eResearch Infrastructures with Adapted Live Streaming Capabilities, Syed Hasan, Laurent Lefevre, Zhiyi Huang and Paul Werstein

Large scale e-Research environments face classical distributed challenges: performance, heterogeneous equipment and variable contexts. The users of such infrastructures want to benefit from full interactive environments based on multimedia streams (voice, video, virtual reality) which are difficult to design and support on a large scale basis. In this paper, we present a new approach to support the streaming of live flows between e-Researchers. We show that traditional techniques (using TCP-based live streaming) are unsuitable for infrastructures with long delay and high loss rate. TCP introduces rate oscillations and requires more buffering and bandwidth to sustain a smooth playback. We propose a streaming framework which provides smoother rate control than TCP and improves streaming performance based on cross-layer feedback between the transport protocol and the streaming server. Our solution keeps the buffer usage at the client and server to a minimum level and provides quick rate adaptation. This paper presents simulation results for streaming in different eResearch scenarios.

Resource Evaluation and Node Monitoring in Service Oriented Ad-hoc Grids, Ian Scriven Andrew Lewis, Matthew Smith and Thomas Friese

Ad-hoc grid computing is an emerging computing technology that promises to deliver high performance at relatively low cost using existing computing resources. There are a number of grid middleware systems being developed to this end. However, a number of features are lacking that are required if ad hoc grid computing is to become viable in a production environment. This paper addresses two of these key features – a resource evaluation and allocation system, which allows grid developers to accurately specify the requirements of their grid job to ensure the most suitable nodes are used when creating the ad-hoc grid, and a node monitoring and error recovery system, which allows grid applications to detect and recover from errors and complete successfully. These systems are built into Mage, the Marburg Ad-hoc Grid Environment; a grid middleware solution developed using the Globus Toolkit, Apache Tomcat and FreePastry.

Grid Resource Allocation: Allocation Mechanisms and Utilisation Patterns, Stefan Krawczyk and Kris Bubendorfer

Grid systems have been put to remarkable use in recent years. Finding planets, rendering multi-million dollar movies, and helping to understand disease are just some of the examples grid systems have been used for. With business turning to towards using grid systems and looking to make them global mechanisms for service delivery, they are nicely poised to be an exciting future prospect. However the performance of a grid system is strongly related to how well grid resource allocation is performed. With many possible approaches to grid resource allocation we have to ask the question, what impact does the choice of resource allocation methodology have on the utilisation and performance of a Grid system? This paper addresses this question through the investigation of the characteristic allocation patterns for three different resource allocation mechanisms and their subsequent effect on resource utilisation within a simulated Grid system.