In this paper, we present cuLib, a R package that provides an easy-to-access interface for utilizing the computing power of NVIDIA GPU. the cuLib package aims to make GPU-based parallel programming easier, flexible an...
详细信息
ISBN:
(纸本)9781509018949
In this paper, we present cuLib, a R package that provides an easy-to-access interface for utilizing the computing power of NVIDIA GPU. the cuLib package aims to make GPU-based parallel programming easier, flexible and high-performance. It allows the use of GPU computing in R without further knowledge because the syntax for definition and manipulation of GPU data is similar to formal R language. cuLib is compatible with device and operation system. the data interface is very flexible, enabling the users manipulate data freely. the cuLib package provides an R wrapper for libraries of NVIDIA's CUDA toolkit and numerous operations. More importantly, it is not only a mathematical tool but also practicable for the algorithm in dealing with data intensive computation.
Withthe increasing popularity of in-memory computing, Spark [1] has been highly successful in implementing large scale data intensive applications, especially for those that reuse data across multiple parallel operat...
详细信息
ISBN:
(纸本)9781509018949
Withthe increasing popularity of in-memory computing, Spark [1] has been highly successful in implementing large scale data intensive applications, especially for those that reuse data across multiple parallel operations. However due to the fact that Moore's Law has slowed down and memory resources are still costly, we presented an elastic data persisting solution for Spark, which enables data compression to save more heap space for JVM and reducing disk I/O throughput for faster data access. We mathematically derived the criteria for selecting the optimal data compression and persisting plan. Our evaluation of the preliminary prototype of this elastic data persisting solution shows that it can provide resource management recommendations by accounting for input data type, memory space and CPU resource, and can consistently yield high performance that accelerates Spark up to 6×.
Hadoop distributed File System (HDFS) provides the storage to keep analyzing outcomes for the diversity of frameworks. MapReduce, Storm, and Spark each applies on batching, streaming and in-memory computing, all of th...
详细信息
ISBN:
(纸本)9781509018949
Hadoop distributed File System (HDFS) provides the storage to keep analyzing outcomes for the diversity of frameworks. MapReduce, Storm, and Spark each applies on batching, streaming and in-memory computing, all of them need the HDFS to collect and assemble results. For coping with Big-Data analysis in the real world, complicated platforms required working together. However, collaborating analysis on heterogeneous frameworks, the data must be write-through firstly and post-fetch upon HDFS that degrades the performance and lower the effectiveness of the whole system. For best our knowledge, no previous work had focused on inter-framework data caching. To solve above problems on collaborating analysis within heterogeneous frameworks such as Hadoop and Strom, in this paper, we propose a cache system upon YARN called " Inter-Framework Cache" (IF-cache). It uses in-memory cache to reserve temporary outcomes while also reducing the HDFS access frequency and improve analysis performance. Experiments had shown that Hadoop with IF-cache can reduce about 50% times comparing the no-cache one.
this paper introduces a scalable solution for distributing content-based video analysis tasks using the emerging MapReduce programming model. Scalable and efficient solutions are needed for this type of tasks, as the ...
详细信息
ISBN:
(纸本)9781467362337
this paper introduces a scalable solution for distributing content-based video analysis tasks using the emerging MapReduce programming model. Scalable and efficient solutions are needed for this type of tasks, as the number of multimedia content is growing at an increasing rate. We present a novel implementation utilizing the popular Apache Hadoop MapReduce framework for both analysis job scheduling and video data distribution. We employ face detection as a case example because it represents a popular visual content analysis task. the main contribution of this paper is the performance evaluation of distribution models for video content processing in various configurations. In our experiments, we have compared the performance of our video data distribution method against two alternatives solutions on a seven node cluster. Hadoop's performance overhead in video content analysis was also evaluated. We found Hadoop to be a data efficient solution with minimal computational overhead for the face detection task.
this special issue of Journal of Physics A: Mathematical and theoretical appears on the occasion of the 5thinternationalsymposium on Quantum theory and Symmetries (QTS5), held in Valladolid, Spain, from 22–28 July ...
this special issue of Journal of Physics A: Mathematical and theoretical appears on the occasion of the 5thinternationalsymposium on Quantum theory and Symmetries (QTS5), held in Valladolid, Spain, from 22–28 July 2007. this is the fith in a series of conferences previously held in Goslar (Germany) 1999, QTS1; Cracow (Poland) 2001, QTS2; Cincinnati (USA) 2003, QTS3; and Varna (Bulgaria) 2005, QTS4. the QTS5 symposium gathered 181 participants from 39 countries working in different fields of theoretical physics. the spirit of the QTS conference series is to join researchers in a wide variety of topics in theoretical physics, as a way of making accessible recent results and the new lines of different fields. this is based on the feeling that it is good for a physicist to have a general overview as well as expertise in his/her own field. there are many other conferences devoted to specific topics, which are of interest to gain deeper insight in many technical aspects and that are quite suitable for discussions due to their small size. However, we believe that general conferences like this are interesting and worth keeping. We like the talks, in both plenary and parallel sessions, which are devoted to specific topics, to be prepared so as to be accessible to any researcher in any branch of theoretical physics. We think that this objective is compatible with rigour and high standards. As is well known, similar methods and techniques can be useful for many problems in different fields. We hope that this has been appreciated during the sessions of the QTS5 conference. the QTS5 conference offered the following list of topics: 1. Symmetries in string theory, quantum gravity and related topics 2. Symmetries in quantum field theories, conformal and related field theories, lattice and noncommutative theories, gauge theories *** computing, information and control 4. Foundations of quantum theory 5. Quantum optics, coherent states, Wigner functions 6. Dynamical and integr
暂无评论