9:00-10:30 First Session: Welcome to PADABS and Tutorials
- Welcome (Vittorio Scarano, Rosario De Chiara and Gennaro Cordasco)
- Tutorial on D-MASON (Vittorio Scarano and Carmine Spagnuolo, Università di Salerno, Italy)
- Tutorial on Repast (Jonathan Ozik, Argonne National Laboratory, USA)
11:00-12:30 Second Session: Parallel and Distributed Agent-Based Simulations (Chair: Vittorio Scarano)
- Alban Rousset, Bénédicte Herrmann, Christophe Lang, Laurent Philippe. A survey on parallel and distributed Multi-Agent Systems
- Abstract: Simulation has become an indispensable tool for researchers to explore systems without recourse to real experiments. Depending on the characteristics of the modeled system methods used to represent the system may vary. Multi-agent systems are, thus, often used to model and simulate complex systems. Whatever type of modelling is used, increasing the size and the precision of the model increase the amount of computation, requiring the use of parallel systems when it becomes too large. In this paper, we focus on parallel platforms that support multi-agent simulations. Our contribution is a survey of existing platforms and their evaluation in the context of high performance computing. We present a qualitative analysis, mainly based on platform properties, then a performance comparison using the same agent model implemented on each platform.
- Paul Richmond. Resolving conflicts between multiple competing agents in parallel simulations
- Abstract: Agents within multi-agent simulation environments frequently compete for limited resources requiring negotiation to resolve 'conflict'. The negotiation process for resolving conflict often relies on a transactional or serial processes which complicates implementation within a parallel simulation framework. This paper demonstrates how transactional events to resolve competition can be implemented within a parallel simulation framework (FLAME GPU) as a series of iterative parallel agent functions. A sugarscape model where agents compete for space and a model requiring optimal assignment between two populations, the stable marriage problem, are demonstrated. The two case studies act as a building block for more general conflict resolution behaviours requiring negotiation between agents in a parallel simulation environment. Extensions to the FLAME GPU framework are described and performance results are provided to show scalability of the case studies on recent GPU hardware.
- Alessandro Pellegrini and Francesco Quaglia. Programmability and Performance of Parallel ECS-based Simulation of Multi-Agent Exploration Models
- Abstract: While the traditional objective of parallel/distributed simulation techniques has been mainly in improving performance and making very large models tractable, more recent research trends targeted complementary aspects, such as the ``ease of programming". Along this line, a recent proposal called Event and Cross State (ECS) synchronization, stands as a solution allowing to break the traditional programming rules proper of Parallel Discrete Event Simulation (PDES) systems, where the application code processing a specific event is only allowed to access that state (namely the memory image) of the target simulation object. In fact with ECS, the programmer is allowed to write ANSI-C event-handlers capable of accessing (in either read or write mode) the state of whichever simulation object included in the simulation model. Correct concurrent execution of events, e.g., on top of multi-core machines, is guaranteed by ECS with no intervention by the programmer, who is in practice exposed to a sequential-style programming model where events are processed one at a time, and have the ability to access the current memory image of the whole simulation model, namely the collection of the states of any involved object. This can strongly simplify the development of specific models, e.g., by avoiding the need for passing state information across concurrent objects in the form of events. In this article we investigate on both programmability and performance aspects related to developing/supporting a multi-agent exploration model on top of the ROOT-Sim PDES platforms, which supports ECS.
14:00-16:00 Third Session: Parallel Agent-Based Simulations (Chair: Rosario De Chiara)
- Gennaro Cordasco, Francesco Milone, Carmine Spagnuolo and Luca Vicidomini. Exploiting D-MASON on Parallel Platforms: A Novel Communication Strategy
- Abstract: Agent-based simulation models are a powerful experimental tool for research and management in many scientific and technological fields.D-MASON is a parallel version of MASON , a library for writing and running Agent-based simulations.In this paper we present a novel development of D-MASON , a decentralized communication strategy which realizes a Publish/Subscribe paradigm through a layer based on the MPI standard. We show that our communication mechanism is much more scalable and efficient than the previous centralized one.
- Jonathan Ozik, Michael Wilde, Nicholson Collier, Charles M. Macal. Adaptive Simulation with Repast Simphony and Swift
- Abstract: We present a general approach for adaptive ABMS, which integrates Repast Simphony’s distributed batch components and the Swift parallel scripting language. Swift is used to launch Repast Simphony simulations on parallel resources, collect the results from those simulations, and generate further simu-lations based on an analysis of the results. In order to demonstrate the benefits and capabilities of this approach, we developed a simulated annealing reference workflow and applied it to a modified Repast Simphony “JZombies” demon-stration model. The workflow was able to successfully and efficiently find areas of the model parameter space that yielded the desired outcomes, as specified by an objective function. The workflow was run on a high-performance cluster, launching 16 concurrent simulated annealing optimization processes, each exe-cuting 100 simulated annealing loops over 16 stochastic model variations – a total of 25,856 adaptive simulation runs, which was completed in 17 minutes. The materials used in the workflow are presented in a Supplemental Materials section to allow replication.
- Masatoshi Hanai, Anthony Ventresque, Kazuyuki Shudo and Toyotaro Suzumura. Towards a Framework for Adaptive Resource Provisioning in Large-Scale Distributed Agent-based Simulation
- Abstract: We propose a new method for efficient utilization of computational resources in distributed agent-based simulation. Efficient utilization of computational resource becomes important for the simulation user as well as resource providers. This is because such computational resources cost more finely and flexibly, called "pay as you go" charge. But previous simulating methods need to provision enough resource to execute all simulation in advance. This brings about excessive provisioning of computing resource because some agent-based simulations such as a traffic simulation have big computational imbalance between each iterating step. To solve the problem, our method provides a new framework for adaptive recourse provisioning during executing the simulation. In our experiments, we have used totally 9 commodity servers and validate that our method achieves to reduce costs to 74% of the previous way in Tokyo traffic simulation.
16:30-18:00 Fourth Session: Theory and Practice (Chair: Gennaro Cordasco)
- Claudio Cioffi. Theoretical Analysis of Amdahl’s Law using the Nabladot Operator
- Abstract: Amdahl’s Law states that execution speedup S is nonlinearly proportional to the percentage P of code that can be parallelized and to the number N of processors. Since P is continuous but N is discrete, traditional multivariate operators based on nabla or del ∇ are applicable only for P, not for N. Moreover, relatively low values of N (bound by Miller’s number 7 ± 2) are common in some cases. Here I apply a novel and exact operator, called “nabladot” and denoted by the symbol ∇·, that is defined for hybrid function such as Amdahl’s Law. The main results show how exact solutions using nabladot differ from traditional approximations using legacy operators that assume continuity even when not warranted. Improvements in precision are inversely proportional to P and N, converging to 0.8 as N → 2.
- Rosario De Chiara, Ferdinando Picarella, Eugenio Sangianantoni, Paolo Servillo, Francesco Zurolo. On the need of parallelism: a model for pedestrian simulations and its parallelization
- Abstract: Agent-Based Models (ABM) are a class of models which, by simulating the behaviours of multiple agents (i.e., independent actions, interactions and adaptation), aims to emulate and/or predict complex phenomena. The “emergence” of such complex phenomena is often computation intensive and requires tools, libraries and frameworks, capable of speeding up and facilitate the design of complex simulations. In this paper we propose an agent-based model for agorezein and we apply it on a real scenario case of agorezein, people strolling around the city centre. During such event tens of artistic light installations are deployed along the city centre and constitute a great attractor for hundreds of thousands of persons who visit the city in the short period of time of the months right before/after Christmas. We describe how we designed the agents behaviours and how we parallelized it, using DistrSteer, for performances reasons.
- Workshop closing