the trust problem of Cloud Services Providers (CSPs) has become one of the most challenging issues for cloud computing. To build trust between Cloud Clients (CCs) and CSPs, a large number of trust evaluation framework...
详细信息
ISBN:
(纸本)9783030389918;9783030389901
the trust problem of Cloud Services Providers (CSPs) has become one of the most challenging issues for cloud computing. To build trust between Cloud Clients (CCs) and CSPs, a large number of trust evaluation frameworks have been proposed. Most of these trust evaluation frameworks collect and process evidence data such as the feedback and the preferences from CCs. However, evidence data may reveal the CCs' privacy. So far there are very few trust frameworks study on the privacy protection of CCs. In addition, when the number of malicious CCs' feedback increases, the accuracy of existing frameworks is greatly reduced. this paper proposes a robust trust evaluation framework RTEF-pp, which uses differential privacy to protect CCs' privacy. Furthermore, RTEF-pp uses the Euclidean distances between the monitored QoS values and CCs' feedback to detect malicious CCs' feedback ratings, and is not affected by the number of malicious CCs' feedback rating. Experimental results show that RTEF-pp is reliable and will not be affected by the number of malicious CCs' feedback rating.
the widespread of GPS embedded devices has lead to a ubiquitous location dependent services, based on the generated real-time location data. this introduced the notion of continuous querying, and withthe aid of advan...
详细信息
ISBN:
(纸本)9783030389611;9783030389604
the widespread of GPS embedded devices has lead to a ubiquitous location dependent services, based on the generated real-time location data. this introduced the notion of continuous querying, and withthe aid of advanced indexing techniques several complex query types could be supported. However the efficient querying and manipulation of such highly dynamic data is not trivial, processing factors of crucial importance should be carefully thought out such as accuracy and scalability. In this study we focus on Continuous KNN (CKNN) queries processing, one of the most well-know spatio-temporal queries over large scale of continuously moving objects. In this paper we provide an overview of CKNN queries and related challenges, as well as an outline of proposed works in the literature and their limitations, before getting to our contribution proposal. We propose a novel indexing approach model for CKNN querying, namely VS-TIMO. the proposed structure is based on a selective velocity partitioning method, since we have different objects with varying speeds. Our structure base unit is a comprised of a non overlapping R-tree and a two dimensions grid. In order to enhance performances, we design a compact multi-layer index structure on a distributed setting, and propose a CKNN search algorithm for accurate results using a candidate cells identification process. We provide a comprehensive vision of our indexing model and the adopted querying technique.
Withthe advent of next-generation sequencing technology, sequencing costs have fallen sharply compared to the previous sequencing technologies. Genomic big data has become the significant big data application. In the...
详细信息
ISBN:
(纸本)9783030389611;9783030389604
Withthe advent of next-generation sequencing technology, sequencing costs have fallen sharply compared to the previous sequencing technologies. Genomic big data has become the significant big data application. In the face of growing genomic data, its storage and migration face enormous challenges. therefore, researchers have proposed a variety of genome compression algorithms, but these algorithms cannot meet the processing requirements for large amount of biological data and high processing speed. this manuscript proposes a parallel and distributed referential genome compression algorithm-Fast Distributed Referential Compression (FastDRC). this algorithm compresses a large number of genomic sequences in parallel under the Apache Hadoop distributed computing framework. Experiments show that the compression efficiency of the FastDRC is greatly improved when it compresses large quantities of genomic data. Moreover, FastDRC leads to the only distributed computing method known to us in the field of genome compression. the source code for FastDRC can be obtained from this link: https://***/GhostCCCathenry/FastDRC.
the two-volume set LNCS 11944-11945 constitutes the proceedings of the 19thinternationalconference on algorithms and Architectures for parallelprocessing, ica3pp 2019, held in Melbourne, Australia, in December 2019.
ISBN:
(数字)9783030389918
ISBN:
(纸本)9783030389901
the two-volume set LNCS 11944-11945 constitutes the proceedings of the 19thinternationalconference on algorithms and Architectures for parallelprocessing, ica3pp 2019, held in Melbourne, Australia, in December 2019.
the two-volume set LNCS 11944-11945 constitutes the proceedings of the 19thinternationalconference on algorithms and Architectures for parallelprocessing, ica3pp 2019, held in Melbourne, Australia, in December 2019.
ISBN:
(数字)9783030389611
ISBN:
(纸本)9783030389604
the two-volume set LNCS 11944-11945 constitutes the proceedings of the 19thinternationalconference on algorithms and Architectures for parallelprocessing, ica3pp 2019, held in Melbourne, Australia, in December 2019.
Molecular dynamics and many similar time-dependent computing tasks are defined as simple state updates over multiple time steps. In recent years, modern supercomputing clusters have enjoyed fast-growing compute capabi...
详细信息
ISBN:
(纸本)9783030389918;9783030389901
Molecular dynamics and many similar time-dependent computing tasks are defined as simple state updates over multiple time steps. In recent years, modern supercomputing clusters have enjoyed fast-growing compute capability and moderate-growing memory bandwidth, but their improvement of network bandwidth/latency is limited. In this paper, we propose a new communication-avoiding algorithmic model based on asynchronous communications which, unlike BSP, records and handles multiple iterative states together. the basic idea is to let computation run in small regular time steps while communications over longer dynamic time steps. Computation keeps checking inaccuracies so that the intervals between communications are small in volatile scenarios but longer when dynamics is smooth. this helps reduce the number of data exchanges via network communication and hence improve the overall performance when communication is the bottleneck. We test MD simulation of condensed covalent materials on the Sunway TaihuLight. For best time-to-solution, the general-purpose supercomputer Sunway TaihuLight performs 11.8K steps/s for a system with 2.1 million silicon atoms and 5.1 K steps/s for 50.4 million silicon atoms. this time-to-solution performance is close to those of state-of-art hardware solution. A software solution using general-purpose supercomputers makes the technology more accessible to the general scientific users.
Deep neural network training is a common issue that is receiving increasing attention in recent years and basically performed on Stochastic Gradient Descent or its variants. Distributed training increases training spe...
详细信息
ISBN:
(纸本)9783030389611;9783030389604
Deep neural network training is a common issue that is receiving increasing attention in recent years and basically performed on Stochastic Gradient Descent or its variants. Distributed training increases training speed significantly but causes precision loss at the mean time. Increasing batchsize can improve training parallelism in distributed training. However, if the batchsize is too large, it will bring difficulty to training process and introduce more training error. In this paper, we consider controlling the total batchsize and lowering batchsize on each GPU by increasing the number of GPUs in distributed training. We train Resnet50 [4] on CIFAR-10 dataset by different optimizers, such as SGD, Adam and NAG. the experimental results show that large batchsize speeds up convergence to some degree. However, if the batchsize of per GPU is too small, training process fails to converge. Large number of GPUs, which means a small batchsize on each GPU declines the training performance in distributed training. We tried several ways to reduce the training error on multiple GPUs. According to our results, increasing momentum is a well-behaved method in distributed training to improve training performance on condition of multiple GPUs of constant large batchsize.
the Five-hundred-meter Aperture Spherical Radio Telescope (FAST), which is the largest single-dish radio telescope in the world, has been producing a very large data volume with high speed. So it requires a high perfo...
详细信息
ISBN:
(纸本)9783030389611;9783030389604
the Five-hundred-meter Aperture Spherical Radio Telescope (FAST), which is the largest single-dish radio telescope in the world, has been producing a very large data volume with high speed. So it requires a high performance data pipeline to covert the huge raw observed data to science data product. However, the existing solutions of pipelines widely used in radio data processing cannot tackle this situation efficiently. the paper proposes a pipeline architecture for FAST based on HDF5 format and several I/O optimization strategies. First, we design the workflow engine driving the various tasks efficiently in the pipeline;second, we design a common radio data storage specification on the top of HDF5 format, and also developed a fast converter to map the original FITS format to the new HDF5 format;third, we apply several concrete strategies to optimize the I/O operations, including chunks storage, parallel reading/writing, on-demand dump, and stream process etc. In the experiment of processing 700 GB of FAST data, the results show that HDF5 based data structure without other optimizations was 1.7 times faster than original FITS format. If chunk storage and parallel I/O optimization are applied, the overall performance can reach 4.5 times as the original one. Moreover, due to the good expansibility and flexibility, our solution of FAST pipeline can be adapted to other radio telescopes.
暂无评论