Other Researchers: A/Prof Steven Blackburn, Dr Eric McCreath, Dr Peter Strazdins, Dr Uwe Zimmer, Dr Rui Yang
Current PhD/MPhil students: Joseph Antony, Warren Armstrong, Muhammad Atif, Jie Cai, Arrin Daley, Ahmed El Zein, Daniel Frampton, Robin Garner, Pete Janes, Jaison Mulerikkal, Andrew Over, Daniel Robson, Michael Still, H'sein Jin Wong
| Manifesto |
Our educational mission is driven by the depth and breadth of the knowledge about computer systems, and the current and future need to utilise multicore processors, clusters - and other manifestations of concurrency - effectively for high performance and increased computing productivity. Our research mission is driven by creating new programming language infrastructure, software development and programming methods, starting from a very strong base in concurrent computer systems, and language support environments.
| Collaborations |
Collaborations within DCS: Various interactions within the group. Links with Ramesh Sankaranarayana on distributed Java project.
Collaborations within rest of CECS: Early collaboration with Rob Mahony on use of FPGAs, which could include other members of DE. Overlap with Rod Kennedy on use of GPUs and Bioengineering.
Collaborations within rest of ANU: Various interactions with people in ANUSF, SMS, RSC and RSBS.
Other Collaborations: Machine Learning Group at NICTA, Paul Coddington at UAdel, various universities, industry and government laboratories in US, Japan and Singapore.
| Current Grants |
- ARC Linkage: LP0774896, Jan 2007-Dec 2009, with US based companies Sun Microsystems and Gaussian Inc., awarded $250k from ARC, “Programming Paradigms, Tools and Algorithms for Electronic Structure Calculations on Clusters of Non-Uniform Memory Access Parallel Processors”, AP Rendell, PE Strazdins, MJ Frisch, GW Trucks, and SC-W See.
- ARC Linkage: LP0669726, July 2006-June 2009, initially with Australian company Alexander Technology but now supported by Intel, $250k ARC funding “Next Generation Grid Enabled Cluster Computers: Performance Optimisation for e-Science”, AP Rendell and PE Strazdins.
- ARC Linkage: LP0669762, July 2006-June 2009, initially with Australian company Alexander Technology but now supported by Platform Computing, 150k, “Accurate Performance Modelling and Prediction of Cluster Computers”, PE Strazdins and PD Coddington
- ARC Discovery: DP0558228, 2005-2007, $233k “Use of Interval Arithmetic and Grid Computing in Computational Molecular Science: Bounding Errors and Locating Global Minima”, AP. Rendell.
- ARC Discovery: DP0666059, 2005-2009, $575k, “Dynamic Cooperative Performance Optimizations”, S.M. Blackburn and KS McKinley.
| Major Facilities |
The group runs a heterogeneous cluster compute environment with nodes that range from a 16 core Sun X4600 system, an 8-way threaded by 8 core UltraSPARC T2, to a number of dual core Athlons. Both Ethernet and Infiniband network technologies are available, and a variety of special purpose systems are under investigation, including the NVIDIA GTX8800, the IBM Cell processor, and Field Programmable Gate Arrays (FPGAs).
| Project Descriptions |
Research in the group is loosely divided into four themes, with a variety of projects within each.
Security, Reliability and Resilient Computing
DCS Staff: Blackburn, McCreathGraduate Students: Frampton, Garner, Still
Distributed Denial of Service Attacks: Denial of Service (DoS) attacks are deliberate attempts by an attacker to disrupt the normal operation of an Internet based service with the goal of stopping legitimate requests for the service from being processed. This project is focusing on Distributed DoS attacks against Internet email servers (known as SMTP servers), with the aim of developing novel ways of detecting and dealing with this problem.
Garbage Collection: Garbage collection is an important tool for reducing or eliminating a number of classes of programming errors and reducing complexity of application code. We have implemented and maintain MMTk, the leading memory management toolkit used by researchers world wide.
Perpetually Available Software Systems: This research proposes to make software more available, correct, and robust by directly addressing errors in deployed software. Specific goals include (1) keeping real-world software running correctly longer, (2) assist developers in fixing errors and deploying fixes, and (3) solutions which are efficient in time and space so that performance is not noticeably degraded.
Computer Architecture and Performance Modeling
DCS Staff: Blackburn, Edwards, Rendell, Strazdins, Yang, ZimmerGraduate Students: Antony, ElZein Frampton, Garner, Over, Robson
Novel Computing Platforms: In recent years the performance of GPUs and specialized gaming engines such as the Playstation 3 (PS3) has been increasing at a much faster rate than that of general purpose CPUs. Current work is focused on use of the NVIDIA GTX8800 and PS3 for statistical machine learning and molecular dynamics applications, while other early work is considering use of field programmable gate arrays (FPGAs). A second focus is on highly multicore/multithreaded systems, such as the UltraSPARC T2 CoolThreads server.
Software Tools for Modern Architectures: Modern computer systems increasingly incorporate multi-core processors and have memory architectures that have non-uniform memory access. We are interested in developing software tools and methods that are appropriate for such machines.
Modern Architectures and Modern Languages: This project involves microarchitectural performance analysis of languages like Java and C# on current hardware. We are interested in both microarchitectural modifications to support directly the needs of modern managed languages, and the design of asymmetric multicores where cores are particularly targeted at helper activities such as JIT compilation and GC.
HPC and Distributed Computing
DCS Staff: Blackburn, McCreath, Sankaranarayana, Strazdins, RendellGraduate Students: Armstrong, Atif, Cai, Daley, Mulerikkal, Wong
Cluster Computing: We are interested in improving the usability and efficiency of cluster systems. Research targets the development of queuing systems that can optimally allocate user processes across a heterogeneous cluster, and use of virtualization to provide compute environments tailored to a given application.
Distributed Computing Environments: One project in this area is concerned with the use of the shared memory OpenMP programming model as an alternative to message passing on cluster systems. A second project involves the development of a high performance distributed Java virtual machine that is tolerant to hardware and software faults, while also having efficient memory utilization.
Next Generation HPC Languages: This project is interested in moving beyond the traditional HPC languages of Fortran and C, to new languages and environments that are better suited to massively parallel compute systems. Current work involves the X10 programming language under development at IBM Research, and use of dynamic code modification as a means of tuning a running application.
Computational Science
DCS Staff: Rendell, YangGraduate Students: Antony, Janes
Interval Arithmetic: This project is using interval arithmetic to study the effects of rounding and truncation errors for a variety of computational science applications. Also of interest are a variety of novel interval algorithms, such as the ability to rigorously determine global minimums.
Chemical and Biological Systems: We are interested in the development and application of large scale computational science applications. Often these calculations run for days on high performance computing platforms. Systems of current research includes the deposition of Ga on Al2O3 surface, the passage of potassium ions through ion channels, and use of fitting basis sets in density functional theory methods.