Cluster Computing最新文献

As the scale of data centers continues to expand, optimizing resource utilization becomes increasingly critical. Employing virtual machine (VM) migration technology to maintain hosts within an appropriate workload range holds substantial promise for enhancing platform resource utilization, workload equilibrium, and energy efficiency. This study endeavors to reframe virtual machine migration as a partition problem and introduces an integrated framework that adeptly evaluate workload status and precisely identifies the optimal migration target, thus mitigating the expenses associated with virtual machine migration. Our framework commences by employing workload prediction to evaluate host status for determining the most opportune timing for migration. Subsequently, we leverage Service Level Agreements (SLA) violation as the optimization objective to ascertain the optimal status threshold, thereby facilitating effective workload partition of the host. Finally, the framework employs multi-dimensional host resource balance as a guide to schedule host migration in diverse areas, ensuring robust resource utilization post-migration. Experimental results show that compared with three benchmark VM allocation algorithms, SESA, PPRG, and ThrRs. Our framework achieves a significant (17%) increase in multidimensional resource utilization across various types of data centers, accompanied by a noteworthy (27%) reduction in SLA violation rate with fewer time consumption and energy expenditure during VM migration.

The extensive deployment of large cloud data centers has led to substantial energy consumption. Energy conservation is a critical concern for cloud service providers seeking to lower their operating costs within their data centers. To address this energy consumption challenge, effective approaches such as VM consolidation and VM migration are essential. These approaches must carefully balance the trade-off between energy consumption and Service Level Agreement Violations (SLAV). In this paper, we propose an energy-efficient VM selection algorithm for VM consolidation and call it the Simultaneous CPU–Ram Utilization Balancer (SCRUB) policy. This algorithm takes into account CPU and RAM utilization while trying to maintain a balance between energy consumption and SLAV. To evaluate the performance of our proposed method, we implemented it using the Cloudsim simulation toolkit and conducted simulations using real-world workload traces from PlanetLab and Google over three different days. The results show that the SCRUB VM selection policy has led to improvements in various metrics, including reduced energy consumption and a decreased number of VM migrations compared to existing VM selection policies. Specifically, it achieved a 16.98% decrease in energy consumption and a 46.42% reduction in the number of migrations for the PlanetLab dataset, and a 10.95% decrease in energy consumption and a 43.96% decline in the number of migrations for the Google dataset compared to the baseline algorithm MMT.

In this paper, aiming at the problem that the balance between exploration and exploitation of traditional Tasmanian devil optimization algorithm is unflexible, and easy to fall into local optimum, an improved Tasmanian devil optimization algorithm (NTDO) based on the sine-cosine strategy of dynamic weighted factors is proposed. The designed method balances the global and local search capabilities of the algorithm, effectively improves the situation that the algorithm falls into local optimum, and integrally improves the optimization performance of the algorithm. Firstly, the good point set theory is used instead of the traditional random method to find the initial individuals, which can render the initial population is more evenly distributed in the search space and the population diversity is improved. Secondly, A sine-cosine strategy based on dynamic weighted factors is proposed to coordinate the global exploration and local optimization capabilities of the algorithm, and enhance the convergence accuracy of the algorithm. Thirdly, since Tasmanian devil is easy to fall into local optimum in the process of hunting prey, a nonlinear decline strategy based on oscillation factor is presented, which increases the search range of the algorithm and improves the ability of the algorithm to jump out of the local optimal value.Finally, 12 evaluation functions, cec2019 and cec2021 test functions commonly used in NTDO and TDO, WOA, DBO, PSO, GWO, DFPSO and PDGWO algorithms are compared and analyzed, and the experimental results show the effectiveness and feasibility of the scheme.

In the modern, ever-shifting global agri-food environment, the topmost concern revolves around securing the safety, quality, and authenticity of agri-food products. Blockchain technology is being seen as a revolutionary solution for dealing with these issues, providing a decentralized and transparent ledger for the tracking of agri-food products. By incorporating global positioning system and navigation systems within blockchain-based traceability solutions amplifies the capabilities for real-time monitoring, security, and trust. This paper proposes a layered architecture for an efficient agri-food traceability system. The data layer, manages interactions between various entities in the supply chain management and generates agri-food product-related data. The blockchain layer manages data via transactions and smart contracts, using the interplanetary file system for secure, decentralized storage. The navigation layer, combines navigation with Indian constellation and global positioning system to offer precise real-time positioning and timing services, enhancing product tracking. This integrated approach not only improves food safety but also supports sustainability efforts by reducing food waste and bolstering consumer trust in the agri-food industry. We implement the proposed system using Remix IDE, MetaMask wallet, and the Sepolia test network, summarizing the deployment analysis. Performance evaluation is conducted using the JMeter simulation toolkit.The proposed framework achieves an average throughput of 329.26 transactions per second, latency of 49.3 ms, and response time of 87.9 ms. We conduct a comparative evaluation of the proposed system with related studies. From this comparative analysis, we observed that our proposed framework has better features than other related works.

Base station (BS) offers data dissemination as a service to IoT smart devices, enabling efficient reprogramming or reconfiguration for their intended activities in post-deployment. Most of the existing IoT data dissemination schemes rely on flooding, leading to the Redundant Broadcast Storm Problem (RBSP), where multiple sensor nodes repeatedly transmit redundant data to neighbours. RBSP elevates network energy consumption and sender congestion in the network. Given that IoT smart devices communicate through open wireless mediums with the internet as a backbone, they are vulnerable to various malicious threats during data dissemination. Intruders may engage in malicious activities and compromise configuration parameters, leading to device failure to execute intended services. This paper proposes a Secure Cloud-Integrated Data Dissemination Protocol (SCIDP) aimed at ensuring the secure dissemination of data within cloud-integrated environments to mitigate RBSP’s impact and enhances security for performing effective reprogramming of sensor devices in IoT. The proposed protocol is implemented by using NS3 simulator with realistic simulation parameters. Simulation results indicate that the proposed protocol enhances energy efficiency by 12%, dissemination effectiveness by 16%, and network lifespan by 16%. Furthermore, the proposed system decreases communication overhead by 11% and computational costs by 9% compared to alternative existing protocols. From the formal security analysis, the proposed system proves that it can withstand against various kinds of security attacks in the network.

Software-defined networking (SDN) is known for its enhanced network programmability and adaptability, but maintaining strong safety precautions to protect against emerging cyber-attacks remains a constant issue. Since SDN has logically centralized control, an attack on the controller might paralyze the entire network. For this reason, intrusion detection is very crucial. Many academics have embraced state-of-the-art techniques to assess and identify these assaults. However, the majority of these approaches lack scalability and accuracy. Moreover, they had difficulties with restricted features, low efficiency, incorrect characteristics, and computing complexity. Therefore, to detect network vulnerabilities in SDN-based IoT networks, we developed a practical deep learning approach based on Res2Net and Elman Recurrent Neural Networks (ERNN) technique as a defense solution to detect security issues in SDN. This framework consists of multiple steps and starts by addressing the dataset’s class imbalance issue with a Data Augmentation Generative Adversarial Network (DAGAN). Next, the Res2net and Enhanced Honey Badger Algorithm (EHBA) are used to extract and select features. This lowers the computational expense and lessens the possibility that the model would be misled by unsuitable and negative characteristics. Finally, an ERNN-based technique is used to detect and classify the intrusions in SDN. After seeing the network assaults, a practical mitigation framework is implemented to mitigate the network attacks. Three SDN IoT-focused datasets, InSDN, IoT-23 and ToN-IoT, are used in an experimental investigation to analyze the proposed framework’s performance. The results of numerous trials show that the proposed method outperforms existing techniques regarding several constraints.

Skin cancer is one of the most dangerous types of cancer due to its immediate appearance and the possibility of rapid spread. It arises from uncontrollably growing cells, rapidly dividing cells in one area of the body, invading other bodily tissues, and spreading throughout the body. Early detection helps prevent cancer progress from reaching critical levels, reducing the risk of complications and the need for more aggressive treatment options. Convolutional neural networks (CNNs) revolutionize skin cancer diagnosis by extracting intricate features from images, enabling an accurate classification of lesions. Their role extends to early detection, providing a powerful tool for dermatologists to identify abnormalities in their nascent stages, ultimately improving patient outcomes. This study proposes a novel deep convolutional neural network (DCNN) approach to classifying skin cancer lesions. The proposed DCNN model is evaluated using two unbalanced datasets, namely HAM10000 and ISIC-2019. The DCNN model is compared with other transfer learning models, including VGG16, VGG19, DenseNet121, DenseNet201, and MobileNetV2. Its performance is assessed using four widely used evaluation metrics: accuracy, recall, precision, F1-score, specificity, and AUC. The experimental results demonstrate that the proposed DCNN model outperforms other deep learning (DL) models that utilized these datasets. The proposed DCNN model achieved the highest accuracy with the HAM10000 and ISIC-2019 datasets, reaching (98.5%) and (97.1%), respectively. These experimental results show how competitive and successful the DCNN model is in overcoming the problems caused by class imbalance and raising skin cancer classification accuracy. Furthermore, the proposed model demonstrates superior performance, particularly excelling in terms of accuracy, compared to other recent studies that utilize the same datasets, which highlights the robustness and effectiveness of the proposed DCNN.

Thermographic inspection is particularly effective in identifying thermal bridges because it visualizes temperature differences on the building’s surface. The focus of this work is on energy-efficient computing for deep learning-based thermal bridge (anomaly) detection models. In this study, we concentrate on object detection-based models such as Mask R-CNN_FPN_50, Swin-T Transformer, and FSAF. We do benchmark tests on TBRR dataset with varying input sizes. To overcome the energy-efficient design, we apply optimizations such as compression, latency reduction, and pruning to these models. After our proposed improvements, the inference of the anomaly detection model, Mask R-CNN_FPN_50 with compression technique, is approximately 7.5% faster than the original. Also, more acceleration is observed in all models with increasing input size. Another criterion we focus on is total energy gain for optimized models. Swin-T transformer has the most inference energy gains for all input sizes ((approx)27 J for 3000 x 4000 and (approx)14 J for 2400 x 3400). In conclusion, our study presents an optimization of size, weight, and power for vision-based anomaly detection for buildings.

This paper proposes a meta-heuristic called Eel and Grouper Optimizer (EGO). The EGO algorithm is inspired by the symbiotic interaction and foraging strategy of eels and groupers in marine ecosystems. The algorithm’s efficacy is demonstrated through rigorous evaluation using nineteen benchmark functions, showcasing its superior performance compared to established meta-heuristic algorithms. The findings and results on the benchmark functions demonstrate that the EGO algorithm outperforms well-known meta-heuristics. This work also considers solving a wide range of real-world practical engineering case studies including tension/compression spring, pressure vessel, piston lever, and car side impact, and the CEC 2020 Real-World Benchmark using EGO to illustrate the practicality of the proposed algorithm when dealing with the challenges of real search spaces with unknown global optima. The results show that the proposed EGO algorithm is a reliable soft computing technique for real-world optimization problems and can efficiently outperform the existing algorithms in the literature.

Some particular shaped datasets, such as manifold datasets, have restrictions on density peak clustering (DPC) performance. The main reason of variations in sample densities between clusters of data and uneven densities is not taken into consideration by the DPC algorithm, which could result in the wrong clustering center selection. Additionally, the use of single assignment method is leads to the domino effect of assignment errors. To address these problems, this paper creates a new, improved density peaks clustering method use the similarity assignment strategy with K nearest Neighbors (IDPC-SKNN). Firstly, a new method for defining local density is proposed. Local density is comprehensively consider in the proportion of the average density inside the region, which realize the precise location of low-density clusters. Then, using the samples’ K-nearest neighbors information, a new similarity allocation method is proposed. Allocation strategy successfully address assignment cascading mistakes and improves algorithms robustness. Finally, based on four evaluation indicators, our algorithm outperforms all the comparative clustering algorithm, according to experiments conducted on synthetic, real world and the Olivetti Faces datasets.