Nowadays, there is an increasing demand in the High-Performance Computing (HPC) community to make use of different public cloud service provider. The question of which cloud provider is superior for a certain application and usage configuration is very important for the successful deployment of HPC application on the cloud. In this paper, we evaluate the performance of HPC applications on Microsoft Azure cloud platform using the well-known NAS parallel benchmarks. These benchmarks are considered as examples of general scientific HPC applications to test the communication performance. Different process allocation strategies are performed in terms of MOPS and Speedup. Our results show that allocating one process per instance achieves higher scalability at the expense of the cost. The results compared with the same results with the same experiments in Amazon platform. We found that Azure platform has better shared-memory communication performance than Amazon platform. In contrast, Amazon is superior to Azure platform in terms of Ethernet bandwidth.
Cloud computing is a popular emerging computing technology that has revolutionized information technology through flexible provisioning of computing resources. Therefore, efforts to develop an effective resource management approach have found that implementing efficient resource sharing among multiple customers that considers power saving, service-level agreements, and network traffic simultaneously is difficult. This paper proposes a practical integrated pipeline that can use various algorithms. The performance of each algorithm is evaluated independently to obtain the combination of algorithms that guarantees a resource-effective cloud data center framework. This integrated resource management pipeline approach would optimize performance based on several key performance indicators, such as power saving, network traffic, and service-level agreements, for either the whole system or the end-user. The performance of the proposed resource management framework was evaluated using a real testbed. The results demonstrated that the proactive double exponential smoothing algorithm prevents unnecessary migrations, the MMTMC2 VM selection algorithm improved the quality of service for end-users and reduced overall energy consumption and network traffic, and the medium-fit placement algorithm provided load balancing between active servers and decreased service level agreement violations. The performance comparison illustrated that the combination of these algorithms was considered to be the best solution toward a dynamic resource-effective cloud data center. Our results showed that energy consumption and the total number of migrations decreased by 16.64% and 49.44%, respectively.
Efficient resource management approaches have become a fundamental challenge for distributed systems, especially dynamic environment systems such as cloud computing data centers. These approaches aim at load-balancing or minimizing power consumption. Due to the highly dynamic nature of cloud workloads, traditional time series and machine learning models fail to achieve accurate predictions. In this paper, we propose novel hybrid VTGAN models. Our proposed models not only aim at predicting future workloads but also predicting the workload trend (i.e., the upward or downward direction of the workload). Trend classification could be less complex during the decision-making process in resource management approaches. Also, we study the effect of changing the sliding window size and the number of prediction steps. In addition, we investigate the impact of enhancing the features used for training using the technical indicators, Fourier transforms, and wavelet transforms. We validate our models using a real cloud workload dataset. Our results show that VTGAN models outperform traditional deep learning and hybrid models, such as LSTM/GRU and CNN-LSTM/GRU, concerning cloud workload prediction and trend classification. Our proposed model records an upward prediction accuracy ranging from $$95.4\%$$ 95.4 % to $$96.6\%$$ 96.6 % .
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.