Event traces are helpful in understanding the performance behavior of parallel applications since they allow the in-depth analysis of communication and synchronization patterns. However, the absence of synchronized cl...
详细信息
Event traces are helpful in understanding the performance behavior of parallel applications since they allow the in-depth analysis of communication and synchronization patterns. However, the absence of synchronized clocks on most cluster systems may render the analysis ineffective because inaccurate relative event timings may misrepresent the logical event order and lead to errors when quantifying the impact of certain behaviors or confuse the users of time-line visualization tools by showing messages flowing backward in time. In our earlier work, we have developed a scalable algorithm that eliminates inconsistent inter-process timings postmortem in traces of pure MPI applications. Since hybrid programming, the combination of MPI and OpenMP in a single application, is becoming more popular on clusters in response to rising numbers of cores per chip and widening shared-memory nodes, we present an extended version of the algorithm that in addition to message-passing event semantics also preserves and restores shared-memory event semantics.
For an agent-based system that aims to simulate a large number of agents under different scenarios, a proper partitioning mechanism is an essential factor to the performance of the system. However, there is still an i...
详细信息
For an agent-based system that aims to simulate a large number of agents under different scenarios, a proper partitioning mechanism is an essential factor to the performance of the system. However, there is still an issue on selecting/ developing an appropriate partitioning mechanism for a specific agent-based simulation system. One way is to implement distributed agent-based systems using multiple partitioning mechanisms and to compare their performance. However this is in general not feasible. In this paper, we propose a generic framework and a simple evaluation process to address the issue of selecting an appreciate partitioning mechanism by simulating the effect of the partitioning mechanism to predict its performance for an agent-based simulation system.
For an agent-based system that aims to simulate a large number of agents under different scenarios, a proper partitioning mechanism is an essential factor to the performance of the system. However, there is still an i...
详细信息
ISBN:
(纸本)9781617386404
For an agent-based system that aims to simulate a large number of agents under different scenarios, a proper partitioning mechanism is an essential factor to the performance of the system. However, there is still an issue on selecting/ developing an appropriate partitioning mechanism for a specific agent-based simulation system. One way is to implement distributed agent-based systems using multiple partitioning mechanisms and to compare their performance. However this is in general not feasible. In this paper, we propose a generic framework and a simple evaluation process to address the issue of selecting an appreciate partitioning mechanism by simulating the effect of the partitioning mechanism to predict its performance for an agent-based simulation system.
A distributed Multi-Agent System (MAS) is often appropriate in solving large-scale experimental problems. RePast is a well-known simulation toolkit for agent-based modeling. In order to extend the scalability of RePas...
详细信息
ISBN:
(纸本)9781605607986
A distributed Multi-Agent System (MAS) is often appropriate in solving large-scale experimental problems. RePast is a well-known simulation toolkit for agent-based modeling. In order to extend the scalability of RePast, one way is to partition the model into a distributed simulation. Interfacing RePast with the High Level Architecture (HLA) is a way to build the distributed system. This has previously been achieved in the HLA-RePast system by using a traditional interfacing approach which is time-consuming in the development process. In this paper, a new approach to interfacing RePast with the HLA is introduced. It is a pioneering attempt to adopt a generic architecture for Commercial Off-The-Shelf (COTS) simulation package interoperability to interface an agent-based simulation package with the HLA. A middleware called DSManager originally designed to support COTS simulation package interoperability, is proven to be able to support interoperability of RePast models in a fast and easy manner. The advantages and tradeoffs, and the performance of the new distributed RePast system are discussed and compared to the HLA-RePast system.
Simulating crowds is a challenging but important problem. There are various methodologies in the literature ranging from macroscopic numerical flow simulations to detailed, microscopic agent simulations. One key issue...
详细信息
ISBN:
(纸本)9781424457717
Simulating crowds is a challenging but important problem. There are various methodologies in the literature ranging from macroscopic numerical flow simulations to detailed, microscopic agent simulations. One key issue for all crowd simulations is scalability. Some methods address this issue through abstraction, describing global properties of homogeneous crowds. However, ideally a modeler should be able to simulate large heterogeneous crowds at fine levels of detail. We are attempting to achieve scalability through the application of distributed simulation techniques to agent-based crowd simulation. distributed simulation, however, introduces its own challenges, in particular how to efficiently partition the load between a number of machines. In this paper we introduce a method of partitioning agents onto machines using an adapted k-means clustering algorithm. We present, validate and use an analysis tool to compare the proposed clustered partitioning approach with a series of existing methods.
This paper presents our ongoing work on modeling agents with human-like decision making and behavior execution capabilities in crowd simulation. We aim to provide a generic framework that reflects the major cognitive ...
详细信息
This paper presents our ongoing work on modeling agents with human-like decision making and behavior execution capabilities in crowd simulation. We aim to provide a generic framework that reflects the major cognitive and physical processes as observed from human behaviors in real-life situations. The design of the framework is based on some basic assumptions and related cognitive theories on human behaviors in various real-life situations. In this paper, the cognitive architecture of our framework is presented, which emphasizes the role of experiences in human's decision making. The paper also briefly describes the design of agent's decision making process and presents a case study to show some results in a crowd simulation scenario.
Commercial-Off-The-Shelf Simulation Packages (CSPs) are widely used in industry to simulate discrete-event models. Interoperability of CSPs requires the use of distributed simulation techniques. Literature presents us...
详细信息
ISBN:
(纸本)9781424457717
Commercial-Off-The-Shelf Simulation Packages (CSPs) are widely used in industry to simulate discrete-event models. Interoperability of CSPs requires the use of distributed simulation techniques. Literature presents us with many examples of achieving CSP interoperability using bespoke solutions. However, for the wider adoption of CSP-based distributed simulation it is essential that, first and foremost, a standard for CSP interoperability be created, and secondly, these standards are adhered to by the CSP vendors. This advanced tutorial is on an emerging standard relating to CSP interoperability. It gives an overview of this standard and presents case studies that implement some of the proposed standards. Furthermore, interoperability is discussed in relation to large and complex models developed using CSPs that require large amount of computing resources. It is hoped that this tutorial will inform the simulation community of the issues associated with CSP interoperability, the importance of these standards and its future.
Simulation is a low cost alternative to experimentation on real-world physical systems. In light of the prevalence of web services, issues relating to distributed simulation using web services have come to the fore in...
详细信息
Nowadays, more researchers in the simulation field are interested in the use of a service-oriented architecture and component-based system in order to design more efficient large-scale distributed simulation systems. ...
详细信息
Nowadays, more researchers in the simulation field are interested in the use of a service-oriented architecture and component-based system in order to design more efficient large-scale distributed simulation systems. A Service-Oriented Architecture (SOA) is a collection of independent services with well-defined interfaces that can be accessed without knowledge of their implementation. A component-based approach separates the whole system into different parts based on their functionalities. In this paper, a framework for component-based distributed simulation on the Grid (SOAr-DSGrid) is introduced. One of the major issues in designing the framework is to have efficient data FLow and control flow in the framework. Data flow describes the way that data moves in the system while control flow determines the order of activities that happen in the system. The rule engine is a critical part of the framework and is responsible for managing the distributed data flow and control FLow. The design and implementation of the rule engine is the major focus of this paper. The rule engine enables the framework to be more scalable as the components within the framework now communicate in a fully distributed manner.
暂无评论