P.M.A. Sloot

Large Scale Simulations of Elastic Light Scattering by a Fast Discrete Dipole Approximation

Simulation of Elastic Light Scattering from arbitrary shaped particles in the resonance region (i.e., with a dimension of several wavelengths of the incident light) is a long standing challenge. By employing the combination of a simulation kernel with low computational complexity, implemented on powerful High Performance Computing systems, we are now able to push the limits of simulation of scattering of visible light towards particles with dimensions up to 10 micrometers. This allows for the first time the simulation of realistic and highly relevant light scattering experiments, such as scattering from human red — or white blood cells, or scattering from large soot — or dust particles. We use the Discrete Dipole Approximation to simulate the light scattering process. In this paper we report on a parallel Fast Discrete Dipole Approximation, and we will show the performance of the resulting code, running under PVM on a 32-node Parsytec CC. Furthermore, as an example we present results of a simulation of scattering from human white blood cells. In a first approximation the Lymphocyte is modeled as a sphere with a spherical inclusion. We investigate the influence of the position of the inner sphere, modeling the nucleus of a Lymphocyte, on the light scattering signals.

Self-organized criticality in simulated correlated systems

In this paper we study the influence of spatio-temporal correlations on the dynamic runtime behavior of the optimistic parallel Time Warp simulation method. By means of Ising spin simulation, we show experimentally that the probability distribution of the number of rolled back events behaves as a power-law distribution over a large range of sub-critical Ising temperatures and decays exponentially for super-critical Ising temperatures. The experimental results indicate that for critical Ising temperatures, where long-range correlations occur, the computational complexity of Time Warp and physical complexity of the Ising spin model are entangled and contribute both to the runtime behavior in a nonlinear way.

Load balancing by redundant decomposition and mapping

Abstract In this paper a new methodology for load balancing parallel processes on parallel systems is proposed. The problem of load balancing is considered to be an NP-hard optimization task. Taking static parallel finite element applications as a case study, the benefits and losses that follow from applying the methodology are studied. It is found that the proposed methodology can be especially useful for load balancing in asymmetric processor topologies, and therefore is of importance for work load balancing in workstation clusters.

Toward Grid-Aware Time Warp

The authors study the adaptation of an optimistic Time Warp kernel to cross-cluster computing on the Grid. Wide-area communication, the primary source of overhead, is offloaded onto dedicated routing processes. This allows the simulation processes to run at full speed and thus significantly decreases the performance gap caused by the wide-area distribution. Further improvements are obtained by employing message aggregation on the wide-area links and using a distributed global virtual time algorithm. The authors achieve many of their objectives for a cellular automaton simulation with lazy cancellation and moderate communication. High communication rates, especially with aggressive cancellation, present a challenge. This is confirmed by the experiments with synthetic loads. Even then, a satisfactory speedup can be achieved, provided that the computational grain of events is large enough.

Time Warp cancellation optimizations on high latency networks

We investigate the performance of the time warp kernel APSIS when running on various communication layers, in particular on a wide-area grid. Several cancellation strategies are tried, among them the lazy cancellation and a little known bulk anti-messages optimization. Our experiments with an Ising spin simulation indicate that the slowdown caused by high latency networks, while significant, is not catastrophic; and that it can be significantly reduced using the lazy cancellation. Experiments suggest that further improvements can be expected if a more elaborate communication infrastructure is put in place.

Task Allocation by Parallel Evolutionary Computing

In this paper we will investigate the applicability of parallel evolutionary algorithms to the task allocation problem?a long standing problem in parallel computing. Three different evolutionary optimization strategies, genetic algorithms, simulated annealing, and steepest descent, are formulated in a parallel generic framework. In order to enhance the performance of the strategies, a number of adjustments that exploit problem specific knowledge is proposed. We adopt a parametric description of static parallel applications. As a consequence, a theoretical analysis of the task allocation solution space can be conducted with a method originating from computational biology. The prediction following from this analysis, i.e., simulated annealing performs optimally on the solution space, is supported by experimental results.

Simulation of Hiërarchical Resource Management for Meta-computing Systems

Optimal scheduling in meta-computing environments still is an open research question. Various resource management (RM) architectures have been proposed in the literature (e.g. [2][13][12]). In the present paper we explore, through simulation, various multi-level scheduling strategies for compound computing environments comprising several clusters of workstations. We study global and local RM and their interaction. The local RM comprises both the cluster management and operating system schedulers. Each level refines the scheduling decisions of the layer above it, taking into account the latest resource information. Our experiments explore conventional strategies like First Come, First Served (FCFS) and Shortest Job First (SJF) at the global RM level. At all levels, the schedulers strive to maintain a good load balance. The unit of load balancing at the global level is the job consisting of one or more parallel tasks; at the local level it is the task. The results of our simulations indicate that, especially at high system loads, the use of a global RM can result in a significant performance gain.

Simulation of gravitational wave detectors

A simulation program that provides insight into the vibrational properties of resonant mass gravitational radiation antennas is developed from scratch. The requirements that are set necessitate the use of an explicit finite element kernel. Since the computational complexity of this kernel requires significant computing power, it is tailored for execution on parallel computer systems. After validating the physical correctness of the program as well as the performance on distributed memory architectures, we present a number of “sample” simulation experiments to illustrate the simulation capabilities of the program. The development path of the code, consisting of problem definition, mathematical modeling, choosing an appropriate solution method, parallelization, physical validation, and performance validation, is argued to be typical for the design process of large-scale complex simulation codes.

Performance prediction of N-body simulations on a hybrid architecture

Hybrid Architectures consist of Special Purpose Devices (SPD) used in conjunction with parallel computers. They have the potential of being a major computing tool for various applications of Computational Science. Mapping the application tasks to the hybrid machine is not trivial, and bottlenecks can arise quite easily. Performance Modeling is a suitable method to tackle this problem. The performance model presented here simulates the activity of GRAPE, a very powerful SPD used in Computational Astrophysics, in conjunction with a general purpose host. We show how our model faithfully reproduces the behaviour of the real hybrid machine. We also present some examples of how our model can be used to predict performance improvements.

Simulating light scattering from micron-sized particles

Employing the combination of a kernel with low computational complexity, implemented on powerful HPC systems, we are now able to push the limits of simulation of light scattering from arbitrary particles towards particles with dimensions up to 10 micrometer. This allows for the first time the simulation of realistic and highly relevant light scattering experiments, such as scattering from human white blood cells, or scattering from large soot — or dust particles. We use the Discrete Dipole Approximation to simulate the light scattering process. In this paper we report on a parallel Fast Discrete Dipole Approximation, and we will show the performance of the resulting code, running under PVM on a 32-node Parsytec PowerXplorer. Furthermore, we present results of a simulation of scattering from a model of a small Human White Blood Cell. This model is the largest possible particle fitting in memory of our parallel computer, and contains 1.1 million dipoles.