Search

Distributed storage systems such as Ceph have been widely adopted, with erasure coding technology being an essential fault-tolerance technique. While ensuring data reliability and security, it significantly reduces the cost of data storage. Due to the computational overhead and encoding latency introduced by the erasure coding process, the data encoding rate is often constrained. To address this issue, an FPGA-accelerated erasure coding encoding scheme in Ceph, based on an efficient layered strategy (FPGA-Accelerated Erasure Coding Encoding in Ceph with an Efficient Layered Strategy, LFEC-Accelerator), is proposed and implemented. This approach takes full advantage of FPGA's parallel computing capabilities to accelerate the erasure coding algorithm at the hardware level. Furthermore, to maximize the utilization of the FPGA controller's resources and ensure that all processing steps are properly managed and scheduled, our approach introduces a hierarchical structure comprising a communication interface layer, task scheduling layer, and hardware acceleration layer. Experimental results indicate that, under the same erasure coding configurations and file sizes, our solution outperforms native Ceph-supported erasure coding libraries such as Jerasure, Clay, Shec and ISA, with an encoding rate improvement of up to 3.04 times. [ABSTRACT FROM AUTHOR]

Accurate and continuous crop phenology information at a regional scale is important for agronomic management and yield estimation. However, detecting continuous crop phenology remains challenging due to the low sensitivity of remote sensing signals to certain phenological stages and the limited of availability remote sensing images. Therefore, this study developed a layered strategy to detect continuous crop phenology. First, a novel Phenology Separability Index (PSI) is established to select features from the Gaussian probability density distribution. PSI quantifies the capability of optical vegetation indexes (VIs), Synthetic Aperture Radar (SAR) signals, and meteorological factors to distinguish between various phenological stages. Then, the multi-temporal sample is established to enhance training sample representation and quantity. Finally, a random forest model is trained using features extracted from multi-temporal samples to improve detection accuracy. This model effectively reduces phenological stage confusion due to redundant features and limited samples. In addition, this study validated its extensibility by mapping crop phenology in the cities of Acheng, Zhaozhou, Lishu, and Buxin and assessed its uncertainty using the Sobol approach. Results indicated that growing degree day has the highest separability among meteorological factors, surpassing both SAR singles and optical VIs. Moreover, the proposed layered strategy was robust, explaining 96% of spatial variation in crop phenology at the regional scale. The accuracy of the layered strategy method (total RMSE = 8.74 days) surpassed that of the multi-temporal sample method (total RMSE = 15.76 days) and the traditional method with a single-temporal sample (total RMSE = 17.21 days). In addition, this study indicated that optical VIs are prone to confuse with the early or late phenological stage of corn, whereas SAR singles are highly sensitive to jointing date.

Autonomous air combat technology of unmanned combat air vehicles (UCAVs) is a hot issue that is currently being studied by various countries, and maneuvering trajectory prediction is an important part of autonomous air combat research. To address the difficulty of maintaining high prediction accuracy and short prediction time simultaneously in maneuvering trajectory prediction, this paper proposes a maneuvering trajectory prediction method that is based on a layered strategy, which combines long-term maneuvering unit prediction and short-term maneuvering trajectory prediction. In long-term maneuvering unit prediction, the complex trajectory is divided into 21 types of maneuvering units using the four characteristics of maneuvering trajectories, and a maneuvering unit library is established. On the basis of the deep echo state network(DeepESN), to capture multiscale prediction input parameters, autoencoder (AE) technology is incorporated. In addition, to increase the prediction accuracy, adaptive boosting (Ada) learning technology is utilized to build a strong predictor, and seven prediction networks are compared. The results demonstrate that the proposed method realizes the highest prediction accuracy. The single-step prediction time is about 0.002 s, which meets the time requirement. In short-term maneuvering trajectory prediction, the long and short-term memory (LSTM) network is analyzed, and the gaussian random walk strategy particle swarm optimization (GWSPSO) algorithm is used to update the internal weights and biases of the network to overcome the problems of “gradient disappearance” and “gradient explosion”, and a data sharing method is proposed for overcoming the no directionality of optimization algorithms. Compared with four traditional networks, the results demonstrate the method that is proposed in this paper performs better. Compared with the sampling time of 0.3 s, the short-term prediction time of 0.05 s can also meet the requirements. Finally, a long- and short-term layered prediction method is used on a group of complex maneuvering trajectories. The results demonstrate that the prediction accuracy is significantly increased and the real-time requirements are satisfied.