Parallel and Distributed Computation
Overview
The central theme of the Parallel and Distributed Computing Research Group is to design and develop open parallel and distributed systems, programming environments, algorithms, and applications that transparently scales and harness resources from enterprise networks to the Internet. The design challenges arise from the distribution of resources across various administrative domains coupled with their availability varying with time. The fundamental design parameters and issues investigated include: security, resource autonomy, uniform interface, robustness, reliability, scalability, economic incentive for cooperation, quality of service, adaptation, utility-based resource allocation, autonomic management, and service-orientation.
The paragraphs below describe the current activities of the group. For more detailed information, and advice about research opportunities, contact the leader of that project.
Cloud Computing and Distributed Systems
Grid and Utility computing models enable the sharing, selection, leasing, and on-demand aggregation of geographically distributed resources such as computers, databases, applications, and instruments. The Gridbus project, a flagship initiative of the Grid Computing and Distributed Systems (GRIDS) Laboratory, has been developing fundamental, next-generation cluster and grid technologies that support a true utility-driven, service-oriented computing. The Gridbus Project is actively investigating a number of research probes that include: visual Grid application development tools for rapid creation of distributed applications, resource management and scheduling, competitive economy-based Grid broker for scheduling distributed data-oriented applications, cooperative economy-based cluster scheduler, Web-services based Grid market directory, Grid Bank, Grid portals, Grid accounting services, and Grid simulation.
Project Leader: Dr Rajkumar Buyya.
More Information: http://www.cloudbus.org/
Project Members: Hussein Gibbins, Akshay Luther,
and Krishna Nadiminti.
Research Students: Martin Placek,
Rajiv Ranjan
Manjuka Soysa,
Anthony Sulistio,
Srikumar Venugopal,
Chee Shin Yeo,
Jia Yu.
Other Collaborators: Alex Barmouta (University of Western Australia),
Professor Wolfgang Gentzsch (Sun Microsystems, USA),
Benjamin Khoo (IBM Global Services),
Dr Rafael Moreno-Vozmediano (Complutense University of Madrid), and
Associate Professor Martin Savior (School of Physics),
Dr Brian Smith (Walter and Eliza Hall Institute for Medical Research),
Associate Professor Chen-Khong Tham (National University of Singapore),
Dr Lyle Winton (School of Physics),
Funding: This project has received funding from the Australian
Research Council, the University of Melbourne, Sun Microsystems,
the Victorian Partnership for Advanced Computing, Singapore Computer
Systems, and Storage Technology Corporation.
Peer- to-Peer Computing
The emergence of massive, autonomous, peer-to-peer (P2P) systems on the Internet is a spectacular phenomenon that has generated a new level of network programming abstraction and presents significant challenges for parallel and distributed computing applications. P2P networks are transient, completely decentralized and potentially disconnected into autonomous sub-networks of activity. Advancing P2P applications from fundamental file sharing towards more general resource sharing, process management, and ultimately towards a P2P operating system, requires significant understanding of P2P algorithms and network programming technologies. As network technology continues to expand into wireless and ad-hoc networking domains, the use of P2P applications becomes increasingly important and complex. This project is addressing open research questions in the following areas: cost optimal topologies, convergence of distributed systems, parallel programming abstractions, collaborative systems, communication complexity, modeling interconnection network properties and emergent Internet structures.
Project Leader: Dr Aaron Harwood.
More Information: http://p2p.csse.unimelb.edu.au/
Project Members: Yi Nutanong,
Minh Truong.
Research Students: Rajiv Ranjans, Scott Douglas, Khaled Samahi, Tau Chen Tamm.
Other Collaborators: Dr Ron Balsys (Central Queensland University).
Funding: This project has received funding via ARC Discovery
projects and the University of Melbourne.
Distributed Data Management
Wide-area networks storing distributed data have become a reality of our lives. To enable access and querying on distributed data, we need some form of data indexing. Conventional centralized indexes are relatively straight forward to implement but they lead to congestion as the rate of requests increases, and they can form a single point of failure. Index replication and caching may be used to offset some of the congestion and reliability issues. But they do not scale well enough to address the needs of wide-area networks. Besides, central indexes are very difficult to update over certain dynamic networks where many mobile nodes connect and disconnect from the network. This project pursues two major research directions in accessing distributed content. First, we can divide the space that the data lies on with a division algorithm and afterwards assign responsibilities to the members of the network. The key issue in this approach is that the division algorithm should be globally known by all the members of the network and hence can be maintained without an all-to-all communication algorithm. Second, we can divide a conventional central index over the nodes of a network so that connectivity is still preserved without replicating the complete index on each node.
Project Leader: Dr Egemen Tanin.
More Information: http://www.cs.mu.oz.au/~egemen/