The interests of the DHPC group are varied, but cluster around how distributed computing systems can be improved and deployed in interesting applications areas. Like everyone else in distributed computing we are wrestling with how to address issues of: latency and bandwidth; scheduling; and and fault tolerance across heterogeneous computing resources.
We have a distributed computing framework - known as DISCWorld which we use as a software testbed for ideas in our research. This is a high level client-server-server middleware for gluing together applications and services.
We are investigating communications technologies including Java RMI and other object transport protocols. We are also looking at ways to improve conventional messaging technologies such as MPI and RPC by layering higher level services and operations on top of them. Our special interest in this area is how systems can be built to work well across long haul (ie high latency) but also high bandwidth, but potentially unreliable networks.
We maintain an interest in benchmarking and assessing performance of computer systems and applications in parallel and distributed environments. Our present focus is on how the performance of Java applications can be improved, from the perspective of distributed applications, core Java Virtual machine performance on various platforms as well as new special purpose applications libraries.
We have built and maintained an archive of satellite imagery since 1996. This archive of some 5 Terabytes of data provides us with an experimental platform for new research in storage technologies and systems. We are interested in how file systems can be integrated with other management software, and how multi medium storage systems (including disks, arrays of disks and tapes and tape robots) can be operated remotely and transparently to applications and users. One of our research developments is software to allow a PC to view files on a remote tape silo (on the other side of Australia) as a very large hard disk - albeit with some bizarre latecies. We are also developing software to organize parallel I/O on such large remote devices.
We have applied serveral of our research project developments to help build deployable systems. We have built an image archive management system for DSTO using Java and CORBA distributed computing technologies. We are working with DEHAA to build a similar system for managing orthophotographic imagery. We are working with the SA State Herbarium to set up a Internet accessable botanical data access system.
See also the ACSys 2 Research Theme on Object Technologies and Systems, which DHPC is still pursuing.
See also our list of ongoing and completed projects.