The proliferation of cloud computing has prompted leading technology providers to establish extensive cloud data center (CDC) infrastructures. Balancing service quality and profitability is a crucial goal due to the intense competition brought about by constantly upgrading user demands. As a result, suppliers are increasingly relying on complex fault-tolerant mechanisms to ensure the reliability of CDC. The traditional rescheduling method is a commonly used approach. However, this often leads to longer completion times or increased compensation, thereby damaging reputation. We investigates operational inefficiencies attributable to virtual machine (VM) failures and failed cloud task (FCT) rescheduling in CDC. Our contribution includes three aspects. First, we present a dynamic classification rule of FCT according to the attribute for the FCT. Second, building upon this rule, we propose a rescheduling mechanism for CDC leveraging dynamic fault-tolerant classification (RMDFC) algorithm to concurrently enhance failure repair rate and minimize compensation overhead. Finally, perform performance tests on the RMDFC algorithm. The failure repair rate of RMDFC algorithm has increased by an average of 3.8 times compared to the comparative algorithms. At the same time, compensation decreased by an average of 58% and profit increased by an average of 9% for cloud service providers in CDCs. The test results show that the RMDFC algorithm exhibits excellent performance.

As the cloud data center (CDC) landscape continues to broaden, CDC resource utilization as the benchmark for assessing scheduling methodologies. Concurrently enhancing both CPU and memory utilization stands as a paramount priority. However, prevalent algorithms tend to solely prioritize CPU utilization while neglecting memory efficiency, ultimately escalating energy expenditure. This article initiates by conducting a comprehensive examination of memory utilization repercussions on CDCs. Subsequently, it facilitates the dynamic clustering of physical machines and virtual machine deployments, ensuring a balanced utilization profile. Furthermore, it introduces the memory priority scheduling algorithm for CDC based on machine learning dynamic clustering algorithm (PMPD). Comparative evaluations against other algorithms underscore the prowess of PMPD in concurrently optimizing CPU and memory utilization, thereby minimizing the number of active PMs and diminishing energy consumption of CDCs.

As the division of labor in the industry becomes more refined, an increasing number of companies are abandoning infrastructure construction and instead moving their operations to cloud data centers (CBDCs). Cloud service providers are responding to the surge in demand by deploying their own CBDCs worldwide. However, the energy consumption and operation costs of these CBDCs vary depending on the region's environment and policies. To mitigate these costs, cloud service providers often employ resource management algorithms. This article conducts a comprehensive analysis of the cross-regional CBDC model, including establishing virtual machine classification rules based on clustering results. Ultimately, this article proposes a low-energy resource classification algorithm for cross-regional CBDCs based on the K-means clustering algorithm (LCKC). The effectiveness of the LCKC algorithm is compared to that of other algorithms, and the results indicate that it reduces energy consumption in cross-regional CBDCs.

The rapid growth of data centers provides cloud users with better choices and services. An increasing number of companies are involved in the construction of data centers ; at the same time, the original cloud service providers are constantly expanding new data centers to meet the needs of users. Therefore, the joint scheduling of multiple data centers has become a difficult problem that cloud service providers must face. Joint scheduling refers to the resource allocation among multiple data centers of a cloud service provider. That is, how can the virtual machine application of cloud users be deployed on the physical machine in multiple data centers. A single data center provisioning strategy can solve the scheduling problem within the data center. However, multiple data centers can make the previous provisioning strategy invalid due to geographical location and other differences. In this paper, the characteristics of different data centers are analyzed and clustered to form a data center consortium. Then, the virtual machine (VM) is classified according to the deadline. As cloud users purchase virtual machine resources from cloud service providers, they are billed by time in most cases. The deadline of the virtual machine refers to the time when the virtual machine ends its service. Then, we analyze the power consumption of the data center and deploy VMs based on the rule of avoiding hotspot deployment of VMs with priority energy consumption. Finally, a cloud service provisioning strategy using data center consortium clustering (PSDC3) is proposed and verified. It can be seen from the results that the algorithm can significantly reduce the number of hotspots and energy consumption of cloud data centers.