摘要 :
Chunk-based deduplication has been widely used in storage systems to save storage space. However, deduplication impairs data reliability due to the inter-file chunk sharing. The loss of shared chunks will make these referenced fil...
展开
Chunk-based deduplication has been widely used in storage systems to save storage space. However, deduplication impairs data reliability due to the inter-file chunk sharing. The loss of shared chunks will make these referenced files inaccessible. Meanwhile, we find that inter-file and highly-referenced chunks are important that need higher reliability assurance, but occupy a small fraction of physical storage. Traditional deduplication systems utilize erasure coding or replication techniques to ensure data reliability. With the growth of shared chunks, promoting the reliability of erasure-coded systems incurs large I/O cost because of the weakness of coding scalability. Although replication is easy to scale, it incurs larger storage overhead. In this paper, we present DARM, a Deduplication-Aware Redundancy Management approach via exploiting deduplication semantics (e.g., inter-/intra-file duplicates, chunk size and reference count) to improve data reliability with low overhead. DARM leverages erasure coding for storing unique and low-referenced chunks to improve both storage reliability and space efficiency, and employs Selective and Dynamic Chunk-based Replication (SDCR) for maintaining inter-file and highly-referenced chunks to enhance storage reliability. Experimental results based on real-world datasets show that DARM reduces storage overhead by up to 43.4% and achieves at most 12.7% reliability improvements over the state-of-the-art schemes.
收起
摘要 :
Chunk-based deduplication has been widely used in storage systems to save storage space. However, deduplication impairs data reliability due to the inter-file chunk sharing. The loss of shared chunks will make these referenced fil...
展开
Chunk-based deduplication has been widely used in storage systems to save storage space. However, deduplication impairs data reliability due to the inter-file chunk sharing. The loss of shared chunks will make these referenced files inaccessible. Meanwhile, we find that inter-file and highly-referenced chunks are important that need higher reliability assurance, but occupy a small fraction of physical storage. Traditional deduplication systems utilize erasure coding or replication techniques to ensure data reliability. With the growth of shared chunks, promoting the reliability of erasure-coded systems incurs large I/O cost because of the weakness of coding scalability. Although replication is easy to scale, it incurs larger storage overhead. In this paper, we present DARM, a Deduplication-Aware Redundancy Management approach via exploiting deduplication semantics (e.g., inter-/intra-file duplicates, chunk size and reference count) to improve data reliability with low overhead. DARM leverages erasure coding for storing unique and low-referenced chunks to improve both storage reliability and space efficiency, and employs Selective and Dynamic Chunk-based Replication (SDCR) for maintaining inter-file and highly-referenced chunks to enhance storage reliability. Experimental results based on real-world datasets show that DARM reduces storage overhead by up to 43.4% and achieves at most 12.7% reliability improvements over the state-of-the-art schemes.
收起
摘要 :
The article is about an innovative method for color designing. Through magnifying the color variance with angular variation effect, we scheme out a new type of vehicle colors, which show different colors depend on light and movement.
摘要 :
The article is about an innovative method for color designing. Through magnifying the color variance with angular variation effect, we scheme out a new type of vehicle colors, which show different colors depend on light and movement.
摘要 :
The goal of clustering is to partition the data points into clusters, such that the data points in the same cluster are similar. Therefore, similarity measure is one of the most critical issues for clustering. In this paper, we pr...
展开
The goal of clustering is to partition the data points into clusters, such that the data points in the same cluster are similar. Therefore, similarity measure is one of the most critical issues for clustering. In this paper, we present a novel similarity measure based on intrinsic dimension, where the local intrinsic dimension of each data point is considered as a new feature to describe the data points, leading to a new type of similarity measure combining the new feature and original features. The main idea is that the data points in the same cluster are expected to have the same intrinsic dimension while they have similar values of the traditional features. The proposed method is evaluated on some artificial data sets and the experiment results illustrate the effectiveness of the proposed similarity measure. Moreover, the segmentation results of natural images based on the proposed similarity measure show that the intrinsic dimension is worthy of being considered as a new feature of the data points in more applications.
收起
摘要 :
The goal of clustering is to partition the data points into clusters, such that the data points in the same cluster are similar. Therefore, similarity measure is one of the most critical issues for clustering. In this paper, we pr...
展开
The goal of clustering is to partition the data points into clusters, such that the data points in the same cluster are similar. Therefore, similarity measure is one of the most critical issues for clustering. In this paper, we present a novel similarity measure based on intrinsic dimension, where the local intrinsic dimension of each data point is considered as a new feature to describe the data points, leading to a new type of similarity measure combining the new feature and original features. The main idea is that the data points in the same cluster are expected to have the same intrinsic dimension while they have similar values of the traditional features. The proposed method is evaluated on some artificial data sets and the experiment results illustrate the effectiveness of the proposed similarity measure. Moreover, the segmentation results of natural images based on the proposed similarity measure show that the intrinsic dimension is worthy of being considered as a new feature of the data points in more applications.
收起
摘要 :
Oxygen regulator was the core component of the Oxygen Supply System (OSS)
which supplied enough oxygen or mix-oxygen to pilots by the oxygen supply laws.
By analyzing the working principle and the control law of the OSS, a new con...
展开
Oxygen regulator was the core component of the Oxygen Supply System (OSS)
which supplied enough oxygen or mix-oxygen to pilots by the oxygen supply laws.
By analyzing the working principle and the control law of the OSS, a new control
circuit and software were designed for the new type aircraft electronic oxygen
regulator (AEOR). Based on the bottom-up principle, the processes from chip
selection to control law optimum were introduced. After these, the methods that
building the experiment system of the AEOR were introduced, and the validation
experiments on the principle were carried out for the AEOR. The ground simulated
experiment for the AEOR was completed according to the actual using circs. By
analyzing the influences of the motor speed, the inlet pressure, the flight altitude, the
aircraft g-load et al. on the capability of the AEOR, the control program was
optimized to adapt different flight states. The experiment results indicated that the
controller with the AVR singlechip could control the AEOR accurately and the
control program had good readability and efficiency. On the aspect of the breathing
resistance and the system undulation, the experiment results of the AEOR were
excelled to the existing technical index. By the comparison between the experimental
results of the AEOR and the aircraft mechanical oxygen regulator (AMOR), the
AEOR showed better performance than the AMOR on the response speed and the
regulation precision. The AEOR achieved the design request for the new aircraft
Oxygen Supply System (OSS). The last, by the analysis of some problems found in
the experiments, such as the block of the electromotor moving, the fluctuation of the
outlet pressure et al., some effective resolves were worked out and it was important
for the design and improvement of the aircraft electronic oxygen regulator (AEOR).
收起
摘要 :
Oxygen regulator was the core component of the Oxygen Supply System (OSS)
which supplied enough oxygen or mix-oxygen to pilots by the oxygen supply laws.
By analyzing the working principle and the control law of the OSS, a new c...
展开
Oxygen regulator was the core component of the Oxygen Supply System (OSS)
which supplied enough oxygen or mix-oxygen to pilots by the oxygen supply laws.
By analyzing the working principle and the control law of the OSS, a new control
circuit and software were designed for the new type aircraft electronic oxygen
regulator (AEOR). Based on the bottom-up principle, the processes from chip
selection to control law optimum were introduced. After these, the methods that
building the experiment system of the AEOR were introduced, and the validation
experiments on the principle were carried out for the AEOR. The ground simulated
experiment for the AEOR was completed according to the actual using circs. By
analyzing the influences of the motor speed, the inlet pressure, the flight altitude, the
aircraft g-load et al. on the capability of the AEOR, the control program was
optimized to adapt different flight states. The experiment results indicated that the
controller with the AVR singlechip could control the AEOR accurately and the
control program had good readability and efficiency. On the aspect of the breathing
resistance and the system undulation, the experiment results of the AEOR were
excelled to the existing technical index. By the comparison between the experimental
results of the AEOR and the aircraft mechanical oxygen regulator (AMOR), the
AEOR showed better performance than the AMOR on the response speed and the
regulation precision. The AEOR achieved the design request for the new aircraft
Oxygen Supply System (OSS). The last, by the analysis of some problems found in
the experiments, such as the block of the electromotor moving, the fluctuation of the
outlet pressure et al., some effective resolves were worked out and it was important
for the design and improvement of the aircraft electronic oxygen regulator (AEOR).
收起
摘要 :
Hard exudates (HEs) are one kind of the most important symptoms of Diabetic Retinopathy (DR). A new method based on background-estimation for hard exudates detection is presented. Firstly, through background-estimation, foreground...
展开
Hard exudates (HEs) are one kind of the most important symptoms of Diabetic Retinopathy (DR). A new method based on background-estimation for hard exudates detection is presented. Firstly, through background-estimation, foreground map containing all bright objects is acquired. We use the edge information based on Kirsch operator to obtain HE candidates, and then we remove the optic disc. Finally, the shape features, histogram statistic features and phase features of the HE candidates are extracted. We use the SVM classifier to acquire the accurate extraction of HEs. The proposed method has been demonstrated on the public databases of DIARETDB1 and HEI-MED. The experiment results show that the method's sensitivity is 97.3 % and the specificity is 90 % at the image level, and the mean sensitivity is 84.6 % and the mean predictive value is 94.4 % at the lesion level.
收起
摘要 :
Hard exudates (HEs) are one kind of the most important symptoms of Diabetic Retinopathy (DR). A new method based on background-estimation for hard exudates detection is presented. Firsdy, through background-estimation, foreground ...
展开
Hard exudates (HEs) are one kind of the most important symptoms of Diabetic Retinopathy (DR). A new method based on background-estimation for hard exudates detection is presented. Firsdy, through background-estimation, foreground map containing all bright objects is acquired. We use the edge information based on Kirsch operator to obtain HE candidates, and then we remove the optic disc. Finally, the shape features, histogram statistic features and phase features of the HE candidates are extracted. We use the SVM classifier to acquire the accurate extraction of HEs. The proposed method has been demonstrated on the public databases of DIARETDB1 and HEI-MED. The experiment results show that the method's sensitivity is 97.3 % and the specificity is 90 % at the image level, and the mean sensitivity is 84.6 % and the mean predictive value is 94.4 % at the lesion level.
收起