Ken Hawick's Projects

See also my list of student projects.

This is an ongoing list of projects past, distant past and present. I have kept it in one form or another since leaving the USA as a diary of past project work. In brief, some of the projects, partners, customers and collaborators I have worked with here at Massey, at Bangor (2000-2003), Adelaide(1996-2000), Bracknell(1995-1996), Syracuse(1993-1995), and Edinburgh(pre 1993) are included.

I moved to Massey University in 2003, and am still establishing new and continued research activities here. My main research interests are in complex systems and artificial life simulations.

I still have links with the DHPC group which I founded in Adelaide. I am still working on cluster computing and various ideas in distributed services such as storage. Some of this work ties in with recent government programs in e-Science. Other longer term interests are in Computational Science - although sometimes "e-Science" and "Computational Science" are seen as the same area.

Present work is mostly described on my research page - including work on complex systems; artificial life and quantum computer simulations.

Some other research/commercial activities involve designing middleware for computational grids of clusters around new Zealand, and for connecting mechatronics and robotic systems together so they can interoperate and utilise distributed processing power.

Some recent work involves use of the Massey University cluster-based supercomputer (Helix) to carry out large scale simulations of various models in statistical mechanics. This work uses the Monte Carlo method and a range of fast and efficient pseudo random number generators to sample partition functions and state spaces of model such as the small world network system, cluster dynamical systems and various growth models. Some additional work I am doing is looking at use of 64-bit random number generators for use of the new 64-bit chip architectures becoming available. Some work also involves use of object oriented techniques for encapsulating parallelism and random number generation. This work continues the cluster computing projects at Adelaide. We have recently (mid-2005) built a 64-bit cluster from Macintosh G5 nodes to experiment with large memory simulations. The machine is appropriately called "Monte".

Another recent area of work is applying small-world network models to mobile and wireless information management systems to calculate optimal area coverage in dynamic environments.

I am also investigating artificial life systems through simulation of computer agents that evolve. I am focussing on statistical mechanical approaches to investigate the role of the existing laws of statistical physics and thermodynamics in describing ALife.

The Distributed and High Performance Computing Group and the DHPC Technical Notes are described in depth on the Adelaide web site which will remain the definitive one at www.dhpc.adelaide.edu.au. My new Tech note series is in the even more general rubric of Computational Science.

Another strand of ongoing work is on pervasive and ubiquitous networks of handheld devices. In Wales we invested in what were then Compaq handhelds using the cut down Windows operating system and the then nascent blutooth and Wireless Internet technologies. In New Zealand we have re investe d in what are now HP devices and are continuing to look at embedded Java applications. We have recently become distracte d with modelling (simulating) networks of these devices - cheaper than buying lots of them. We re-engineered our "Distributed Information systems Control World" software to build networks of multi threaded daemons that can communicate over ad-hoc networks with various caching strategies.

During my period as Head of Computer Science at Bangor, we ran the DHPC group as a collaboration between Adelaide, Australia and Bangor, Wales. This worked well and Heath James and I had 17 graduate students at one point working in Bangor on various projects. When Heath and I came out to Massey in New Zealand in 2003, we realised that it would be easier to amalgamate all the DHPC material in opne site - rather than leave DHPC web-sites as a trail behind us. Consequently the Adelaide site is now the definitive one and Paul Coddington now looks after the DHPC "brand".

Some of the Bangor-specific projects are described below with continuations noted:

• Applying distributed computing ideas and algorithms to build a swarm of cooperating robots. This work built upon our past distributed systems control software and Java daemons and is using Tini processor boards. The work started as a way of experimenting with very small cluster systems and with the enginnering background of several students who joined as PhD candidates it was natural to involve embedded systems technology. One particular success was building a swarm of "Cybot" robots with modified electronics and embedded Java software for controlling the mobile cluster. The work continues at Bangor.

• We continued our work on cluster computing following the success of the Beowulf clusters we designed and built at Adelaide for colleagues in Chemistry and Physics. We built a prototype Sun Netra cluster for experimentation in fault tolerance and distributed storage algorithms. An additional cluster systems for experiments in routing and advanced networks and new applied systems for fluid dynamics and agricultural simulation and modelling was built and is still in use at Bangor. This work led to a Bangor campus-wide grid with various resources in other department s hooked up for data and cycle sharing. A particularly useful outcome was data used for a bioinformatics grid project grant application.

• We are (still) working on algorithms for Distributed Information Systems Control World - a new architecture for a scalable and robust wide-area computational grid. This work involves algorithm development in namespace management; resource discovery; secure transactions and wide-area data sharing. Distributed Information Systems Control World was originally an attempt to keep me sane while supervising 7 PhD projects in metacomputing. The field blossomed into grid computing and the project ended up taking on a life of its own. My students managed to build several systems to study scheduling, naming, security and other fundamental issues in grid computing. Our original intention was that we would switch to using the emerging Globus system for grid experiments. That turned out to be useful for building some applications prototypes but as Globus grows in complexity it is now very hard for a student to master enough of it in a short enough time to be able to work on some of the fundamental ideas. This is a shame because there are still several interesting grid issues that are worthy of research effort. Consequently Heath James and I are rebuilding the system as a small lightweight student kit that allows experimentation with ideas. We are retrofitting it to use soem of the emerging new protocols and interoperability standards such as WSDL and OGSA.

• We continued some of our JavaGrande work and used Java as a vehicle for parallel and high performance computing systems. Recent work shows how Java makes a powerful glue for configuring clusters and parallel systems in a way that is interoperable with computational grid systems. See Adelaide DHPC site.

• We set up an experimental Virtual Reality project with research focussing on distributed data sharing and collaboration. The system started out as just a couple of workstations with 3d Shutter glasses and a 3d mouse wand. We used this to convince various partners in industry and Welsh government that the ideas could scale and we moved on to using a Sun Blade 200 workstation running Java3D software for visualising simulation configurations. The system was successfully used by our 2001 Summer School students and subsequent honours projects. The work continues and Bangor now has a more elaborate virtual reality facility used for collaborations and simulations in chemistry.

These were all fun projects and involved our group of grad students at Bangor (some MPhil and some PhD). As at Adelaide we managed to create an ongoing portfolio of undergraduate-sized projects to tackle aspects of the bigger projects.

In addition to DHPC, I also worked with Prof Tim Porter in our Mathematics Division in establishing a Logic and Computer Science (LACS) Group to provide theoretical underpinnings for our other computing projects. At present the group is studying logic models; network models and scheduling algorithms and formalisms - all with the aim of furthering the underpinnings of a global computational grid. Bangor has subsequently made the decision to close its Mathematics department which is jeopardising this collaboration.

I also worked with my colleague Heath James in establishing a technology transfer centre. We set up the Centre for Advanced Software Technology (CAST) as a loose confederation of industry collaborations and projects and the Welsh Development Agency (WDA) is now helping us with plans for a custom building for CAST on the Parc Menai business park near Bangor. This project will allow Bangor to run many more industry collaboratons and will have facilities including: spacious machine room capacity; development laboratories; training facilities; short and long term business incubator units; Internet cafe and project planning facilities and a two storey Virtual Reality unit.

This project was funded through the European Union's Objective One programme. Building work is continuing at the time of writing (early 2004).

Connected with CAST we ran a portfolio of industry collaboration projects. Partly these were funded directly by the Welsh Development Agency (WDA) which sponsored my Chair in Computer Science. In addition we obtained funding through the Teaching Company Scheme. This scheme was highly successful in promoting realisable collaboration between our deprtment an Industry. Typically the project would be a joint proposal between the academic and industry partner involved and woudl pay for a graduate student to work at the company while studying part time.

In addition the scheme also provided a fund for academic support on each project. I used the funds accrued from that to support other students in the research group. TCS has now been superceded by Knowledge Transfer Partnerships. Thanks to the efforts of Bangor's Industrial Programme Unit Bangor was at one point number 3 in the UK for success in gaining TCS projects. Heath and I had a total 7 TCS projects operating between us during the period 2000-2003. They included:

• Financial Modelling for Derivatives Trading
• Wide Area Data Grid for Medical Services
• M-Commerce for Veterinary Supplies
• Use of Java to Replace Green Screen APIs in the Retail Industry
• Use of Graph Theory to Model Document Management in the Nuclear, Services, Insurance and Retail Industries.

Last and by no means least, I have spent a not inconsiderable portion of my time since 2000 in setting up the Computer Science teaching programmes at Bangor. We were very pleased in 2002 to have a cohort of eight students graduate from the new BSc in Computer Science and to have the British Computer Society accredit the programme. Student numbers for this programme have been approximately doubling each year since its inception.

We also launched a set of MSc programmes in Distributed Computing; Computational Science; Internet Computing; and also a Conversion MSc for Computing to allow science students from any discipline to train in state of the art computational science. I remain interested in Computational Science in its broadest sense as an emerging new discipline in its own right.

• The Distributed and High Performance Computing Project (DHPC) which I led at Adelaide until April 2000 was a broad project involving many student and commercial sub projects to further the general objective of building metacomputing middleware. In the course of 1996 and subsequently, metacomputing transformed into what is now known as grid computing. Computational grids are systems of wide area distributed computing allowing remote access to powerful computing resources. DHPC had three primary projects in grid technolgies:
  • The Distributed Information Systems Control World Project - a middleware research exploration and prototyping project.
  • The On Line Data Archives Project - an early data grids project.
  • The Distributed Geographic Information Systems Project - grids of geographic data - see also our recent journal article in Parallel Computing Vol 10, PP1297 (2003), describing this work.

  Another important area of DHPC work was in cluster computing. I rescued 8 486 PCs from a junk shop to build our first Beowulf cluster system and the data and student activity we generate d from that launched collaborations with Physics, Chemistry and Geography departments. My interests in cluster suystems continued at Bangor and here at Massey.

  Something of a high risk experiment at the time was to investigate the use of the then emerging VR technology. After some preliminary investigations of the then nascent headtracker technology and a visit to the CAVE facilities and research group in Urbana Illinois and to the National University of Singapore CAVE facility, we started some work into data visualisation and driving for high performance simulations. In collaboration with the South Australian Parallel Computing Centre (Francis Vaughan at Adelaide University) we had some student projects to rig up OpenGL code for visualising output from 3D simulations. In particular all the embarassingly written (but fast) simulations I had used in my PhD work. We ended up with a rather impressive VR room with a back-projected screen, black cloth and paint everywhere to improve the contrast, and a cunningly folded light path with a large mirror to maximise user space. The haberdashery shop in Rundle Mall probably still thinks it weird that Francis and I wanted to buy their entire stock of black cloth. Some architecture applications and a set of game-style demos proved a draw for students and this facility bootstrapped thinking and ideas that led to the designs for the embedded VR facility in the CAST building later in Wales.

  These are all described at the DHPC Web site at Adelaide, project pages.

• Helping Prof Geoffrey Fox establish the electronic InfoMall - an early attempt at a virtual organisation. The Northeast Parallel Architectures Center (NPAC) at Syracuse University was itself part of a technology driven virtual organization - the Center for Research on parallel Computation, directed by Prof Ken Kennedy at Rice University and involving Rice, Syracuse, Caltech, Argonne and Tennessee. InfoMall became a flagship industry outreach programme - with over 60 companies involved at various levels. NPAC, CRPC and InfoMall pioneered the use of broadband ARTM networks to demonstrate the potential for virtual organisations. The project was at the height of its importance when Hilary Clinton visited Syracyuse in 1994 to see a demonstration of the network in telemedecine.

• MultiMedia Information Systems for the USAF as part of the InfoMall technology transfer program. This work helped the US Air Force and other parts of the US Department of Defence make use of the new fangled Web which was then in its infancy. The project involved investigating various embryo web technologies for their applicability in wide area information management across the Air Force.

• Computational Electromagnetics Codes for Syracuse Research Corporation. This project used a data-parallel approach and some very fast dense-matrix solvers running on the Connection Machine and compute clusters to compute radar signatures of various flying hardware components.

• A Survey of High Performance Computing Systems , as part of the US National HPC Software Exchange (NHSE).

• Building and Evaluating High Performance Fortran Applications. When HPF was a relatively new idea, I led an applications-oriented project to build demonstrator applications and use these to evaluate the efficiency and capabilities of HPF systems.

• The Department of Defense High Performance Computing Modernization Programme - whereby I spent a huge ammount of time buried in Nichols Research Corporations' proposal writing facility. Nicholls' successful consortium bid incorporated a Programme for Education and Training (PET) which I helped design. The Programme (including its PET component) ran successfully for several years, transferring high performance computing skills and technology across the US Department of Defence.

• The UK Meteorological Office's Unified Weather and Climate Model code. It needed to be ported to a parallel computing environment - a project that started in 1990 and took until 1996 to complete. This lives on to the present as a whole new generation of parallel computing people in the Met Office now operate the T3E massively parallel system. The work started when I led the numerical simulations group at Edinburgh Parallel Computing centre. EPCC had collaborative contracts with the Met Office to investigate how the unified weather and climate model could be parallelised on the various parallel architectures of the day. I was involved with this work until 1995 when joined the Met Office at Bracknell, to help port the ocean component of the model to a parallel systems.

• Developing Fingerprint Recognition software parallelisation for the UK Home Office. This work looked at the encoding of fingerprints as compact features and the comparison of "prints" and scene of crime "marks" could be compared very quickly using their feature-space representations on a parallel system so that a particular mark could be rapidly checked agains entire national or indeed international databases of known prints and marks. Our project looked at parallel Fortran codes for doing this on the Connection Machine and other systems.

• Black Oil Reservoir Simulation code for Intera. The EPCC project involved porting it to a distributed computing environment. The main difficulty was the sheer size of the code - 250,000 lines of Fortran. A directives-based approach to introducing paralleism eventually led to a parallel product version of the code.

• Building Parallel Computing Benchmarks for the National Physical Laboratory. In the early 1990's parallelism was becoming mainstream and the NPL became involved in setting up some benchmarks to assist companies and government with procurement. I led EPCC's contract project to help NPL.

• Computational Fluid Dynamics Codes for the UK Atomic Energy Authority to figure out how liquid sodium flows around reactor cooling systems. This work used a message passing approach to an existing Fortran code.

• Computational Fluid Dynamics Codes for Rolls Royce to build better aeroengines - well to stop the turbo blades cracking anyway. This Fortran code was ported to run on one of the earliest cluster computer systems deployed in industry - making use of lost cycles on desktop workstations.

• Probablisitic Safety Assessment codes for UK Nirex to tell them where it was safe to bury nuclear waste. These simulation codes needed to be sped up to allow an increasing number of what-if scenarios to be evaluated.

• My own PhD project was on "Domain Growth in Alloys". I was supervised by Prof Stuart Pawley at Edinburgh and Prof Colin Windsor at the Atomic Energy Authority Research Establishment, Harwell. I looked at the dynamic scaling of domains at they grew in various alloys including FeCr, NiMo and CuCo. I also looked at strange growth in the PE16 steel alloy - an important problem at the time as nuclear fuel tie-bars were made of this material. I spent time doing neutron scattering experiments at Harwell's Pluto reactor and at the Intense Pulsed Neutron Source at the Argonne National Laboratory in the USA. I also did a number of parallel computer simulations on the Distributed Array Processors (DAP) at Edinburgh and on the Edinburgh Concurrent Supercomputer transputer system. In the course of all this British Telecom paid me to help set up their research transputer system at their Martlesham Heath research laboratories in 1988. I was also lucky enough to have the UK's Alvey Committee pay me to report on high performance computing activities at Argonne and at Caltech. I also visited Caltech as a research visitor, so it was rather a busy 3 year PhD. I wrote up the final stages of my thesis while working as a project manager in Prof David Wallace's Edinburgh Parallel Computing Centre (EPCC) in 1990.