Keynote Speakers of the DCABES 2007

 

 

Novel Distributed Computing Techniques for Mobile Telecommunications

ABSTRACT  This paper presents a novel distributed technique based on the application of InfiniBand technology over the RDMA channel Interface to develop a distributed Service Level Agreement (SLA) and enforcement solution for the generalised mobile messaging infrastructure. The paper reports on the construction of a distributed SLA application that is able to handle extremely high throughput and still provide the required data integrity. We developed an SLA quota management model, deployed across distributed servers that provide a comparison illustrating the aftermath of deploying the SLA enforcement solution over InfiniBand and Ethernet technologies respectively.

Keywords: SLA (Service Level Agreement), InfiniBand, RDMA (Remote Direct Memory Access), MPI (Message Passing Interface), Distributed Systems.

 

 

Scheduling Parallel Processes and Load Balancing
In Large-scale Computing Systems

Prof. V.P. Kutepov

Applied Mathematics Department, Moscow Power Engineering Institute (Technical University)

ABSTRACT  In the paper the problem of the development effective methods and operating tools for management of large-scale computing systems is discussed. In particular scheduling parallel processes and load balancing strategies and algorithms implemented in our projects on software for computing systems are presented.

 

Keywords: Computing systems, Parallel programming, Scheduling parallel processes, Managing worklo

Vitaliy Pavlovich Kutepov is a Doctor of Technical Science and an Honored Professor, the Vice-Chair of the Applied Mathematics Department at Moscow Power Engineering Institute (Technical University) (MPEI) He graduated from MPEI in 1961 with the specialty “Computer Science”. In 1968 he obtained the Ph.D. on the theme “Research of Some Formal Methods of the Description and the Optimization of Computing Processes”. In 1982 he won the Doctor's Degree on the theme “Functional Systems and Parallel Computing”. From 1987 to 2003 he was the Head of the Applied Mathematics Department in MPEI. He was visitor of Mathematical and Computing Center of Amsterdam (1972), Department of Computer Science at Ediburgh University (1986), Department of Computer Science at Manchester University (1992). He is one of the founders of network MPEI, and the head of “the Center of Supercomputer Technology”. He has published over 100 scientific journal papers and books, and trained 22 doctors of technical science. His research interests are in the theory of parallel programming, the management of parallel processes in large computer systems, the logical models and languages, the theory of the directed relation, the functional models and languages of parallel programming and the architecture of computer systems.

 

 

Laplace Transform Time Domain-decomposition
For Diffusion Problems

Prof. A J Davies

School of Physics, Astronomy and Mathematics, University of Hertfordshire

Hatfield, Hertfordshire, AL10 9AB, UK
Email: a.j.davies@herts.ac.uk  

ABSTRACT In most processes for the solution of parabolic diffusion problems the time derivative is handled using a finite difference approach. An alternative approach is to use the Laplace transform in time to obtain an elliptic problem in the transform space. The resulting problem may be solved using any appropriate elliptic solver. The Laplace transform approach provides a natural time domain-decomposition for diffusion problems and may be used for both linear and non-linear problems.

Keywords: Laplace transform, time domain-decomposition, diffusion problems.

 

 

Scalable Parallel Algorithms for Multi-component Problems

Prof. Xiao-ChuanCai

Department of Computer Science, University of Colorado at Boulder
Boulder, CO80309, USA
Email: cai@cs.colorado.edu

Abstract Scalability is one of the most important issues in parallel computations when the size of the problem is large and when the number of processors is large. Domain decomposition methods are very useful for the partitioning of a large problem into many independent sub problems and for solving the problems on large scale parallel computers. The scalability of the methods are well-studied for scalar elliptic equations. In this work we investigate the methods for solving the coupled nonlinear system of equations arising from the discretization of multi physics problems.

 

 

On Transformation Methods and Concurrency in Time Domain Computation

Prof. Choi-Hong Lai

Department of Mathematical Sciences, University of Greenwich, London SE10 9LS, UK

ABSTRACT:  The paper examines various transformation methods and their roles in the parallelization of time integration of an unsteady problem in science and engineering. Many engineering and applied science problems required the solutions of nonlinear diffusion equations where the nonlinear feature usually comes with the material properties or the conductivity.  In the case of unsteady problems a time-marching scheme, usually with time step length restrictions, is employed in any temporal integration procedure. These restrictions are usually due to stability criteria of an explicit scheme or the truncation errors of an implicit scheme in approximating the temporal derivatives.  Computing time of such numerical methods inevitably becomes significant.  On the other hand fine grain parallelization of time stepping becomes difficult and it is almost impossible to achieve a distributed/parallel algorithm that is able to yield a de-coupling of the original problem.  There are also many problems which require solution details not at each time step of the time-marching scheme, but only at a few crucial steps and the steady state.  Therefore effort in finding fine details of the solutions using many intermediate time steps is considered being wasted.  Such effort becomes significant in the case of nonlinear problems where a linearisation process, which amounts to an inner iterative loop within the time-marching scheme, is required.  It would be a significant save in computing time when the linearisation process and the time-marching scheme can both be done in parallel.  The main objective of the present work is to remove the time stepping and to combine it with parallel/distributed computers.

Keywords: Non-linear problems, Parallel and Distributed Computing, Time marching.

Choi-Hong Lai is Professor of Numerical Mathematics and head of Scientific Computing and Algorithms Group, School of Computing and Mathematical Sciences, University of Greenwich, UK. His research interest is parallel numerical methods for partial differential equations and their applications in science and engineering.  He is the editor of the Journal of Algorithms and Computational Technology, Multi-Science Publishing, UK, the dedicated journal for DCABES post-conference publication, and has edited many special issues for various international journals.  He is a Visiting Professor at Southern Yangtze University and an Adjunct Professor at Fuzhou University.