Francisco Borges

Improving Communication Patterns for Distributed Cluster-based Individual-oriented Fish School Simulations

Parallel discrete event simulation (PDES) have shown to be an useful paradigm for simulating complex and large-scale models. An individual-oriented approach allows modelers capture complex emerging global behaviors generated by simple local interaction, like observed in self-organized systems. Usually, this type of simulations are highly expensive in terms of computing and communications. One one hand, we can reduce the computing involved in individual interactions by means of developing a robust partitioning method. On the other hand, we have to be able to efficiently handle a huge number of individuals interacting with other individuals stored in memory of remote processors. In this work we will analyze and compare three communication strategies: synchronous and asynchronous message passing (via MPI) and bulk-synchronous parallel (BSP) for our distributed cluster-based individual-oriented fish school simulator. In this type of simulations, the main contributions of our work are: a) we showed that distributed time-driven simulations do not always improve the performance when using synchronous communication strategies, b) we show asynchronous communications strategies are more efficient. In addition, we have verified that the bulk-synchronous parallel method is a scalable.

Strip Partitioning for Ant Colony Parallel and Distributed Discrete-event Simulation☆

Data partitioning is one of the main problems in parallel and distributed simulation. Distribution of data over the architecture directly influences the efficiency of the simulation. The partitioning strategy becomes a complex problem because it depends on several factors. In an Individual-oriented Model, for example, the partitioning is related to interactions between the individual and the environment. Therefore, parallel and distributed simulation should dynamically enable the interchange of the partitioning strategy in order to choose the most appropriate partitioning strategy for a specific context. In this paper, we propose a strip partitioning strategy to a spatially dependent problem in Individual-oriented Model applications. This strategy avoids sharing resources, and, as a result, it decreases communication volume among the processes. In addition, we develop an objective function that calculates the best partitioning for a specific configuration and gives the computing cost of each partition, allowing for a computing balance through a mapping policy. The results obtained are supported by statistical analysis and experimentation with an Ant Colony application. As a main contribution, we developed a solution where the partitioning strategy can be chosen dynamically and always returns the lowest total execution time.

Crowd Evacuations SaaS: An ABM Approach☆

Abstract Crowd evacuations involve thousands of persons in closed spaces. Having knowledge about where the problematic exits will be or where the disaster may occur can be crucial in emergency planning. We implemented a simulator using Agent Based Modelling able to model the behaviour of people in evacuation situations and a workflow able to run it in the cloud. The input is just a PNG image and the output are statistical results of the simulation executed on the cloud. This allows to provide the user with a system abstraction and only a map of the scenario is needed. Many events are held in main city squares, so to test our system we chose Siena and we fit about 28,000 individuals in the centre of the square. The software has special computational requirements because the results need to be statistically reliable. Because these needs we use distributed computing. In this paper we show how the simulator scales efficiently on the cloud.

Optimal Run Length for Discrete-event Distributed Cluster-based Simulations

Abstract In scientific simulations the results generated usually come from a stochastic process. New solutions with the aim of improving these simulations have been proposed, but the problem is how to compare these solutions since the results are not deterministic. Consequently how to guarantee that the output results are statistically trusted. In this work we apply a statistical approach in order to define the transient and steady state in discrete event distributed simulation. We used linear regression and batch method to find the optimal simulation size. As contributions of our work we can enumerate: we have applied and adapted the simple statistical approach in order to define the optimal simulation length; we propose the approximate approach to normal distribution instead of generate replications sufficiently large; and the method can be used in other kind of non-terminating science simulations where the data either have a normal distribution or can be approximated by a normal distribution.

A Hybrid MPI+OpenMP Solution of the Distributed Cluster-based Fish Schooling Simulator

Abstract Exploring the multi-core architecture is an important issue to obtaining high performance in parallel and distributed discrete-event simulations. However, the simulation features must fit on parallel programming model in order to increase the performance. In this paper we show our experience developing a hybrid MPI+OpenMP version of our parallel and distributed discrete- event individual-oriented fish schooling simulator. In the hybrid approach developed, we fit our simulation features in the following manner: the communication between the Logical Processes happens via message passing whereas the computing of the individuals by OpenMP threads. In addition, we propose a new data structure for partitioning the fish clusters which avoid the critical section in OpenMP code. As a result, the hybrid version significantly improves the total execution time for huge quantity of individuals, because it decreases both the communication and management of processes overhead, whereas it increases the utilization of cores with sharing of resources.

Care HPS: A high performance simulation tool for parallel and distributed agent-based modeling

Abstract Parallel and distributed simulation is a powerful tool for developing complex agent-based simulation. Complex simulations require parallel and distributed high performance computing solutions. It is necessary because their sequential solutions are not able to give answers in a feasible total execution time. Therefore, for the advance of computing science, it is important that High Performance Computing (HPC) techniques and solutions be proposed and studied. In literature, we can find some agent-based modeling and simulation tools that use HPC. However, none of these tools are designed to enable the HPC expert to be able to propose new techniques and solutions without great effort. In this paper, we introduce Care High Performance Simulation (HPS), which is a scientific instrument that enables researchers to: (1) develop techniques and solutions of high performance distributed simulations for agent-based models; and, (2) study, design and implement complex agent-based models that require HPC solutions. Care HPS was designed to easily and quickly develop new agent-based models. It was also designed to extend and implement new solutions for the main issues of parallel and distributed solutions such as: synchronization, communication, load and computing balancing, and partitioning algorithms. We conducted some experiments with the aim of showing the completeness and functionality of Care HPS. As a result, we show that Care HPS can be used as a scientific instrument for the advance of the agent-based parallel and distributed simulations field.

Crowd Turbulence with ABM and Verlet Integration on GPU Cards

Managing crowds is a key problem in a world with a growing population. Being able to predict and manage possible disasters directly affects the safety of crowd events. Current simulations focus mostly on navigation, but crowds have their own special characteristics and phenomena. Under specific conditions, a mass can turn into a crowd turbulence which may lead to a possible disaster. Understanding the internal phenomena is an important issue in order to model behavior. In the particular case of crowd turbulence, agents are moved by the crowd by a series of pushes, an involuntary movement that can be hard to reproduce. We propose a simple model to implement this complex problem based on intentional and involuntary interactions among the agents. The implementation is a hybrid model between the Verlet integration method and Agent Based Modeling. We implemented the proposed model using C and OpenCL and we evaluated its performance on a Nvidia GPU.

An agent-based model for assessment of aedes aegypti pupal productivity

Dengue is a febrile disease whose main vector transmitter is the Aedes Aegypti mosquito. This disease has an annual register of 50 million infections worldwide. Simulations are an important tool in helping to combat and prevent the epidemic and, consequently, save lives and resources. Therefore, in this paper, we propose an Agent-Based Model for assessment of the pupal productivity of the Aedes Aegypti mosquito. In this model, the reproduction of the mosquito takes into account the productivity of each type of container. The preliminary results show the effects of considering the pupal productivity for the control and prevention of dengue. As a result, we observed that the prevention methods must consider pupal productivity and that the distance between containers might leverage productivity and increase transmission risk. We verify the completeness and functionality of the model through experimentation using Netlogo.

Crowd Evacuation Modeling and Simulation Using Care HPS

Abstract The problem of evacuating crowded closed spaces, such as discotheques, public exhibition pavilions or concert houses, has become increasingly important and gained attention both from practitioners and from public authorities. This kind of problem can be modeled using Agent-Based Model techniques and consequently simulated in order to study evacuation strategies. In this paper, we show the Fira of Barcelona evacuation model implemented with Care HPS. As a main contribution: i) we extend and added new partitioning approaches and other features in Care HPS to carry on this model; and ii) we figured out that crowd evacuation problem has bottlenecks in reality, such as exits, that required more deep optimization in code in order to decrease the total execution time. Finally, we draw some conclusions and point out ways in which this work can be further extended.

Crowd dynamics modeling and collision avoidance with OpenMP

This paper deals with the problem of simulating crowd evacuations in multicore architectures. Manage evacuations is an important issue that involves lives, and in the case of crowds, thousands of lives. We present our model, able to hold thousands of agents walking through the scenario avoiding dynamic and static obstacles. We test how the model behaves with agents falling down, becoming an obstacle. Simulators are widely used in the area of crowd evacuations. In the present work we introduce a crowd model and a HPC simulation tool. We used OpenMP to program a multicore architecture.