In this paper, we present a system-level fault identification algorithm, using a parallel genetic algorithm, for diagnosing faulty nodes in large heterogeneous systems. The algorithm is based on a probabilistic model ...
详细信息
ISBN:
(纸本)1932415610
In this paper, we present a system-level fault identification algorithm, using a parallel genetic algorithm, for diagnosing faulty nodes in large heterogeneous systems. The algorithm is based on a probabilistic model where individual node fails with an a priori probability p. The assumptions concerning test outcomes are the same as in the PMC model, that is, fault-free testers always give correct test outcomes and faulty testers are totally unpredictable. The parallel diagnosis algorithm was implemented and simulated on randomly generated large systems. Simulations results are provided showing that the parallel diagnosis did improve the efficiency of the evolutionary diagnosis approach, in that it allowed faster diagnosis of faulty situation, making it a contribution to present techniques.
Studies to extend Java have been investigated for widespread use and reuse of high-performance computing resources. Despite the benefits achieved, applications and computation entities are still closely bound to each ...
详细信息
ISBN:
(纸本)1932415262
Studies to extend Java have been investigated for widespread use and reuse of high-performance computing resources. Despite the benefits achieved, applications and computation entities are still closely bound to each other making it difficult to share resources. To allow applications to share computation entities, a generic infrastructure for distributed computing is established in this paper. This infrastructure provides a scalable concurrent computation environment, where the concepts of relay engine and computation routing are developed to coordinate client computation requests, scheduling of computation entities, and other computation resources. The phase 0 implementation in Java verified that it worked as expected and provided scalability for varying arrival rates of computation requests.
Significant additions have recently been made to the occam concurrent programming language. The new occam-pi now supports, among other features, mobile channels, mobile processes, shared channels, and dynamic forking ...
详细信息
ISBN:
(纸本)1932415610
Significant additions have recently been made to the occam concurrent programming language. The new occam-pi now supports, among other features, mobile channels, mobile processes, shared channels, and dynamic forking of new concurrent processes. These features should greatly enhance the ability of occam to precisely and easily implement complex concurrent applications. We have recently evaluated these additions by building a distributed combinatorial optimization algorithm of moderate size (3,000 lines). occam-pi allowed us to quickly put the system together, with relatively little effort. The concurrent features of the implementation have been verified in CSP. The resulting code is easy to extend and can be made fault-tolerant.
Though it is known that clos interconnection networks have many advantages over the other interconnection networks, it would be interesting to see the performance benefits that clos networks can bring to the real worl...
详细信息
ISBN:
(纸本)1932415610
Though it is known that clos interconnection networks have many advantages over the other interconnection networks, it would be interesting to see the performance benefits that clos networks can bring to the real world applications. In this paper, we compare the performance of Fast Fourier Transforms (FFT) in clos networks with that of in mesh networks, a popular interconnection topology. We use both two- dimensional (2-d) and three-dimensional (3-d) formulations of FFT for our studies. We show that the performance of FFT in clos networks is significantly better and we further show that 3-d formulation outperforms the 2-d formulation.
From the very beginning of the CUDA technology, it was essential to apply state-of-the-art optimization techniques. Only then was it possible to fully utilize the enormous computational power of graphic processing uni...
详细信息
ISBN:
(纸本)9781665435741
From the very beginning of the CUDA technology, it was essential to apply state-of-the-art optimization techniques. Only then was it possible to fully utilize the enormous computational power of graphic processing units. However, with the development of the CUDA architecture, the impact of typical optimization techniques on software performance has changed significantly. This article shows how the impact of several optimization techniques on the performance of the image filtering algorithm has changed for the subsequent generations of CUDA architecture. Then, based on the results obtained, it attempts to answer whether tedious and time-consuming optimization of the CUDA software is still necessary.
Object-oriented databases provide rich abstraction facilities for the representation of complex application objects together with their properties and behaviour. Query processing in these databases is quite different ...
详细信息
ISBN:
(纸本)1932415262
Object-oriented databases provide rich abstraction facilities for the representation of complex application objects together with their properties and behaviour. Query processing in these databases is quite different from the processing of the corresponding relational databases. Pairs of application objects having at least a connection (relation) between their members can be interpreted as conceptual binary relations. Object-oriented queries can be processed handling uniformly these relations using the binary relational algebra, which provides an algebraic framework [7] that is analogous [8], [10] to the relational algebra for relational databases [1],[2]. However, query-processing efficiency cannot be easily achieved in these very large object-oriented databases using traditional methods and techniques. In such advanced databases the system's performance can be improved using wavefront algorithms and parallelprocessingtechniques. In this paper wavefront array processors are investigated for the purpose of object pair composition, which takes place during the query-processing phase in object-oriented databases that their functionality is based upon binary relational algebra.
Collaborative applications usually register user interaction in the form of semi-structured plain text event log data. Extracting and structuring this data is a prerequisite for later key processes such as the analysi...
详细信息
ISBN:
(纸本)1932415610
Collaborative applications usually register user interaction in the form of semi-structured plain text event log data. Extracting and structuring this data is a prerequisite for later key processes such as the analysis of interactions, assessment of group activity, or the provision of awareness and feedback. Yet, in real situations of online collaborative activity the processing of log data is usually done offline since structuring event log data is, in general, a computationally costly process and the amount of log data tends to be very large. techniques to speed and scale up the structuring and processing of log data with minimal impact on the performance of the collaborative application are thus desirable in order to be able to process log data in real time. In this paper we present a parallel Grid-based implementation for processing in real time the event log data generated in collaborative applications. Our results show the feasibility of using grid middle-ware to speed and scale up the process of structuring and processing semi-structured event log data. Our Grid prototype follows the Master-Worker paradigm, is implemented using the Globus Toolkit 3.2 and is tested on the Planetlab platform.
File-sharing applications remain today the most representative and popular realization of the Peer-to-Peer paradigm. Large objects receive an increasing amount of interest in such systems. In this paper, we identify s...
详细信息
ISBN:
(纸本)1932415262
File-sharing applications remain today the most representative and popular realization of the Peer-to-Peer paradigm. Large objects receive an increasing amount of interest in such systems. In this paper, we identify several challenges related to sharing voluminous content such as movies, OS distributions, games, etc, in unstructured Peer-to-Peer networks. We then describe our scheme which adaptively expands or contracts system resources in order to improve the sharing process and achieve a fair load distribution among the providers.
In this paper, we propose a new addressing scheme for hexagonal interconnection networks. This is the first addressing system, using two coordinates, assuming both the unicity of the coordinates for each node and a di...
详细信息
ISBN:
(纸本)1892512459
In this paper, we propose a new addressing scheme for hexagonal interconnection networks. This is the first addressing system, using two coordinates, assuming both the unicity of the coordinates for each node and a distance formula between two nodes. Then, we present an optimal routing algorithm based on the distance formula in order to use shortest paths from the source to the destination node.
This paper presents a dynamic load balancing mechanism used in MPIAB (Message Passing Interface - Agent Based), an agent based architecture for parallel programming. The design of the dynamic load balancing module pro...
详细信息
ISBN:
(纸本)1932415610
This paper presents a dynamic load balancing mechanism used in MPIAB (Message Passing Interface - Agent Based), an agent based architecture for parallel programming. The design of the dynamic load balancing module provides automatic determination of the efficient hosts while tasks are being executed to optimize the overall system-wide performance. This is achieved by monitoring the work load of the participating hosts in short time intervals and providing a mechanism for the migration of the agents (tasks) from overloaded hosts to under utilized systems while maintaining the connections among the participating agents with less overhead.
暂无评论