In the field of the cross media systems, one of the important issues is how to extract and deal with the impression of media data to realize impression-based media retrieval. In this paper, we present impression extra...
详细信息
ISBN:
(纸本)9780889867901
In the field of the cross media systems, one of the important issues is how to extract and deal with the impression of media data to realize impression-based media retrieval. In this paper, we present impression extraction and correlation calculation methods for automatic cross media synchronization as a new approach for automatic VJ (Visual Jockey) systems with impression calculation. The significant features of our methods are: 1) cross media impression extraction from media data (music and image), 2) correlation calculation in impression between cross media, such as impression of music and impression of an image, for VJ applications, 3) application to an experimental VJ system. This paper shows several experimental results to clarify the feasibility and effectiveness of our framework and methods.
Recently, data storage systems with data deduplication have been introduced as a method of reducing storage space by eliminating redundant data. In a deduplication storage system, the collision-resistant fingerprint o...
详细信息
ISBN:
(纸本)9780889868113
Recently, data storage systems with data deduplication have been introduced as a method of reducing storage space by eliminating redundant data. In a deduplication storage system, the collision-resistant fingerprint of each data segment must be calculated using a hash algorithm. This paper presents a GPU based accelerator, called g-Dedu, for processing the hash computation of the deduplication system. The g-Dedu accelerator algorithm is especially designed for handling the variable and small size of the data used in a deduplication system, which cannot be processed efficiently by a GPU in a straightforward way. Our data organization approach uses a hierarchical data structure to organize the processing data. A scheduler manages these data for optimal GPU processing. Our patterned data segment approach overcomes some noticeable performance drops resulting from the GPU memory model. Furthermore, different from some previous GPU hash accelerator work, our approach strictly follows the hash processing standard. Using this new approach, g-Dedu achieves 6 times speedup on the SHA-1 computation, and 7.4 times speedup on the SHA-2 computation when compared with a CPU-based mplementation.
In this paper we present the performance of our parallel multi-frontal direct solver when applied to solve linear systems of equations resulting from discretizations of a hp Finite Element Method (hp-FEM). The hp-FEM ...
详细信息
ISBN:
(纸本)9780889867901
In this paper we present the performance of our parallel multi-frontal direct solver when applied to solve linear systems of equations resulting from discretizations of a hp Finite Element Method (hp-FEM). The hp-FEM generates a sequence of computational meshes delivering exponential convergence of the numerical error with respect to the mesh size (number of degrees of freedom). A sequence of meshes is obtained by performing several hp refinements starting from an arbitrary initial mesh. The solver constructs initial elimination tree for an arbitrary initial mesh, and expands the elimination tree each time the mesh is refined. The solver has been tested on 3D Direct Current (DC) borehole resistivity measurement simulations problems. We compare the solver with two versions of the MUMPS parallel solver: with (1) distributed entries executed over the entire problem, and (2) the direct sub-structuring method with parallel MUMPS solver utilized to solve the interface problem. We show that by providing to the solver the knowledge about the structure of the hp-FEM, the order of elimination is obtained straightforward, and leads to a better performance than by submitting the entire matrix to the solver and executing a connectivity graph based ordering algorithm.
The Peer-to-Peer (P2P) networks is heavily used for content distribution applications and are becoming increasingly popular for Internet file sharing. Generally the download of a rite can take from minutes up to sever...
详细信息
ISBN:
(纸本)9781424429271
The Peer-to-Peer (P2P) networks is heavily used for content distribution applications and are becoming increasingly popular for Internet file sharing. Generally the download of a rite can take from minutes up to several hours depending on the level of network congestion or the service capacity fluctuation. In this paper, we consider two major factors that have significant impact on average download time, namely, the spatial heterogeneity of service capacities in different source peers and the temporal fluctuation in service capacity of a single source peer. We prove that both spatial heterogeneity and temporal correlations in service capacity increase the average download time in P2P networks and then analyze a simple, distributed algorithm to minimize the file download time. Here, we have designed a new distributed algorithm namely Dynamically distributedparallel Periodic Switching (D2PS) that effectively removes the negative factors of the existing parallel downloading, chunk based switching, periodic switching, thus minimizing the average download time. There are two schemes (i) parallel Permanent Connection, and (ii) parallel Random Periodic Switching in our Dynamically distributedparallel Periodic Switching (D2PS) method. In our parallel Permanent Connection, the downloader randomly chooses multiple source peers and divides the file randomly into chunks and download happens in parallel for the fixed time slot I and source selection function does not change for that fixed time slot. In our parallel Random Periodic Switching, the downloader randomly chooses multiple source peers and divides the file randomly into chunks and download happens in parallel for each randomly selected time slot t and source selection function changes for each time slot. Our Dynamically distributedparallel Periodic Switching (D2PS) effectively removes correlations in the capacity fluctuation and the heterogeneity in space, thus greatly reducing the average download time.
The proceedings contain 37 papers. The topics discussed include: a model for personalized communications control in pervasive systems;class algebra;an R-tree collision detection algorithm for polygonal models;a method...
ISBN:
(纸本)9780889867901
The proceedings contain 37 papers. The topics discussed include: a model for personalized communications control in pervasive systems;class algebra;an R-tree collision detection algorithm for polygonal models;a method to assist the maintenance of object-oriented software systems using impact analysis;user preference management in a pervasive system should be a trusted function;multi-level processor scheduling using ant optimization;constraint based tutor generator for intelligent tutoring system;SVDD based probabilistic ranking scheme for information retrieval;flowable service model for seamless integration of services;checking protocol conformance in component models using aspect oriented programming;real-time multiversion repeatable read isolation level;and using the volunteer computing XTREMweb-CH: lessons and perspectives.
Conventional autotuning configuration of parameters in distributedcomputing systems using evolutionary strategies increases integrated performance notably, though at the expense of consuming too much measurement time...
详细信息
ISBN:
(纸本)9781424449231
Conventional autotuning configuration of parameters in distributedcomputing systems using evolutionary strategies increases integrated performance notably, though at the expense of consuming too much measurement time. An ordinal optimization (OO) based strategy is proposed in this work, combined with neural networks to improve system performance and reduce measurement time, which is fast enough to autotune configurations for distributedcomputing applications. The method is compared with a well known evolutionary algorithm called Covariance Matrix Algorithm (CMA). Experiments are carried out using high dimensional rastrigin functions, which show that OO can reduce one to two orders of magnitude of simulation time while at the cost of an acceptable scope of optimization performance. We also carried out experiments using a real application system with three-tier web servers. Experimental results show that OO can reduce 40% testing time on average at a reasonable and slight cost of optimization performance.
This paper attempts to identify the reliability of detection of licensed primary transmission based on cooperative sensing in cognitive radio networks. With a parallel fusion network model, the correlation issue of th...
详细信息
ISBN:
(纸本)9780769536101
This paper attempts to identify the reliability of detection of licensed primary transmission based on cooperative sensing in cognitive radio networks. With a parallel fusion network model, the correlation issue of the received signals between the nodes in the worst case is derived. Leveraging the property of false sensing data due to malfunctioning or malicious software, the optimizing strategy, namely Fault-Tolerant algorithm for distributed Detection (FTDD) is proposed, and quantitative analysis of false alarm reliability and detection probability under the scheme is presented. In particular, the tradeoff between licensed transmissions and user cooperation among nodes is discussed. Simulation experiments are also used to evaluate the fusion performance under practical settings. The model and analytic results provide useful tools for reliability analysis for other wireless decentralization-based applications (e.g., those involving robust spectrum sensing).
To reduce the energy consumption in wireless sensor networks during target tracking, a distributed target tracking method applied in asynchronous wireless sensor networks was proposed. Firstly, the dynamic clusters we...
详细信息
ISBN:
(纸本)9780769535593
To reduce the energy consumption in wireless sensor networks during target tracking, a distributed target tracking method applied in asynchronous wireless sensor networks was proposed. Firstly, the dynamic clusters were constructed and adjusted according to the distances between nodes and target, which were the unit for time computing. The cluster headers were responsible for was calculation and transferring of tracking time among different clusters. Then, the set of particle was separated to some subsets by parallel particle filter, which was sampled, weighed and resampled in several nodes. Finally, the estimation of local states was implemented by cluster header through gathering the results uploaded from each node. The simulation results show that parallel particle filter has good performances on tracking accuracy and can reduce communication traffic about 38% compared with center particle filter.
Execution of applications on upcoming high-performance computing (HPC) systems introduces a variety of new challenges and amplifies many existing ones. These systems will be composed of a large number of "fat&quo...
详细信息
ISBN:
(纸本)9780769535449
Execution of applications on upcoming high-performance computing (HPC) systems introduces a variety of new challenges and amplifies many existing ones. These systems will be composed of a large number of "fat" nodes, where each node consists of multiple processors on a chip with symmetric multithreading capabilities, interconnected via high-performance networks. Traditional system software for parallelcomputing considers these chip multiprocessors (CMPs) as arrays of symmetric multiprocessing cores, when in fact there are fundamental differences among them. Opportunities for optimization on CMPs are lost using this approach. We show that support for fine-grained parallelism coupled with an integrated approach for scheduling of compute and communication tasks is required for efficient execution on this architecture. We propose Phoenix, a runtime system designed specifically for execution on CMP architectures to address the challenges of performance and programmability for upcoming HPC systems. An implementation of message passing interface (MPI) atop Phoenix is presented. Micro-benchmarks and a production MPI application are used to highlight the benefits of our implementation vis-A-vis traditional MPI implementations on CMP architectures.
Since the amount of information is rapidly growing, there is an overwhelming interest in efficient network computing systems including Grids, public-resource computing systems, P2P systems and Cloud computing. In this...
详细信息
ISBN:
(纸本)9783642043932
Since the amount of information is rapidly growing, there is an overwhelming interest in efficient network computing systems including Grids, public-resource computing systems, P2P systems and Cloud computing. In this paper we take a detailed look at the problem of modeling and optimization of network computing systems for parallel decision tree induction methods. Firstly, we present a comprehensive discussion Oil mentioned induction methods with a special focus on their parallel versions. Next, we propose a generic optimization model of a network computing system that can be used for distributed implementation of parallel decision trees. To illustrate our work we provide results of numerical experiments showing that the distributed approach enables significant improvement of the system throughput.
暂无评论