Applied Mechanics and Materials Vols. 519-520

Abstract: With the explosive growth of data, the performance of querying the huge amounts of data is more and more important.Simply Method of improving the performance of stand-alone already can't meet the demand.We need to handle huge amounts of data with distributed technology.However, traditional ways are just outrageous abuse of distributed resources, there are a lot of methods to improve efficiency and throughput.In terms of saving resources effectively, index is an indispensable tool.This article embarks from the traditional indexing mechanism, based on the advantage of the traditional methods in combination with the advantages of bitmap index, put forward a kind of extended local index mechanism,which is effective in reducing the traditional local indexes for distributed resources unnecessary overhead.At the same time it also improve the query efficiency. Finally, the test results show that the improved algorithm is efficient.

Abstract: Distributed search techniques of Hadoop are researched and analyzed. Combined with Lucene indexing objects, a search engine system IS successfully built. Efficiency of the system in both time and space is investigated. Merit of distributed processing architecture for a single architecture in data handling is verified. The access and update of file information in distributed search technology are further explored. The research plays a positive role in promoting study of related fields

Abstract: In the light of the excellent distributed storage and parallel processing feature of hadoop cluster, a new kind of network public opinion classification method based on Naive Bayes algorithm in hadoop environment is studied. The collected public opinion documents are stored locally according to the HDFS architecture, and whose character words are extracted paralleled in Mapreduce process. Thus the naive Bayesian classification algorithm is parallel encapsulated on cloud computing platform. The MapReduce packaged Naive Bayesian classification algorithm performance is verified and the results show that the algorithm execution speed are significantly improved compared to a single server. Its public opinion classification accuracy rate is more than 85%, which can effectively improve the classification performance of network public opinion and classification efficiency.

Abstract: The explosive growth of information to promote the development and evolution of the network storage system structure, object-storage system came into being, the object-based storage system stores the data in a storage device with intelligent in Object-based Storage Device, and the use of the metadata server to manage the metadata, but when the metadata access amount is very large, there is a potential performance bottleneck exist in MDS, so metadata server load balancing is very important, this paper studies the metadata server load balancing technology, proposed a dynamic equilibrium strategy based on permissions.

Abstract: As statistics from different software are heterogeneous and cannot be shared, an ontology model-SO^sRCE based on Integrated System for health Information of Peoples Police (abbreviated to PPHIIS) was investigated. The method was combined with agents cooperation. Meanwhile, the model not only solved the problem of semantic integration in PPHIIS but also enabled medical staff to extract detailed medical information more efficiently.

Abstract: Aiming at storing and transmitting the real time data of energy management system in the industrial production, an online data compression technique is proposed. Firstly, the auto regression model of a group of sequence is established. Secondly, the next sampled data can be predicted by the model. If the estimated error is in the allowable range, we save the parameters of model and the beginning data. Otherwise, we save the data and repeat the method from the next sampled data. At Last, the method is applied in electricity energy data compression of a beer production. The application result verifies the effectiveness of the proposed method.

Abstract: This paper presents an overview of a traffic simulation system model for parallel processing in multi-computers on a local network. It serves as the basis for the development of a micro-simulation system for a traffic network in a large-scale. The article also introduces the time synchronization issue, which needs to be solved when executing parallel simulation systems and application of multi-threads programming technique to build a program for the system.

Abstract: To investigate the performance of acceleration technologies for FDK algorithm, two of the most common high-performance computing hardware, multi-core CPU and GPU, are involved in our experiment. Both runtime and accuracy are regarded as the standards to evaluate the performance of four different programming methods: OpenMP, GLSL, CUDA and OpenCL. All the methods are estimated with comparable optimization strategies. The experimental results show that GPU has higher efficiency than multi-core CPU for fast cone-beam reconstruction, meanwhile CUDA is the best choice for programming on the multi-processor featured GPU.

Abstract: With the arrival of the intermodality era, to design fast and efficient K shortest paths (KSP) algorithms becomes gradually one of the core technologies in traveler information systems. Yen is a classical algorithm for KSP. However, Yen is time-consuming. In view of powerful general-purpose computing capabilities, GPU(Graphics Processing Units) has received increasing attention in various fields. Based on CUDA software development environment, combined with the structure of the Yen algorithm itself, this paper proposed two parallel algorithms for Yen. The first parallel algorithm computes candidate shortest paths for very possible deviation nodes in parallel. The second one computes candidate shortest paths in serial, but computes very candidate path in parallel. Finally, the efficiency of the two parallel algorithms was tested through experiments.

Abstract: Dissolved gas analysis (DGA) in oil is an important method for transformer fault diagnosis. This paper use random forest parallelization algorithm to analysis the dissolved gases in transformer oil. This method can achieve a fast parallel fault diagnosis for power equipment. Experimental results of the diagnosis of parallelization of random forest algorithm with DGA samples show that this algorithm not only can improve the accuracy of fault diagnosis, and more appropriate for dealing with huge amounts of data, but also can meet the smart grid requirements for fast fault diagnosis for power transformer. And this result also verifies the feasibility and effectiveness of the algorithm.