Future computation of cloud datacenter resource usage is a provoking task due to dynamic and Business Critic workloads. Accurate prediction of cloud resourceutilization through historical observation facilitates, eff...
详细信息
Future computation of cloud datacenter resource usage is a provoking task due to dynamic and Business Critic workloads. Accurate prediction of cloud resourceutilization through historical observation facilitates, effectively aligning the task with resources, estimating the capacity of a cloud server, applying intensive auto-scaling and controlling resource usage. As imprecise prediction of resources leads to either low or high provisioning of resources in the cloud. This paper focuses on solving this problem in a more proactive way. Most of the existing prediction models are based on a mono pattern of workload which is not suitable for handling peculiar workloads. The researchers address this problem by making use of a contemporary model to dynamically analyze the CPU utilization, so as to precisely estimate data center CPU utilization. The proposed design makes use of an Ensemble Random Forest-Long Short Term Memory based deep architectural models for resource estimation. This design preprocesses and trains data based on historical observation. The approach is analyzed by using a real cloud data set. The empirical interpretation depicts that the proposed design outperforms the previous approaches as it bears 30%-60% enhanced accuracy in resourceutilization.
As an alternative centralized systems, which may prevent data to be stored in a central repository due to its privacy and/or abundance, federated learning (FL) is nowadays a game changer addressing both privacy and co...
详细信息
As an alternative centralized systems, which may prevent data to be stored in a central repository due to its privacy and/or abundance, federated learning (FL) is nowadays a game changer addressing both privacy and cooperative learning. It succeeds in keeping training data on the devices, while sharing locally computed then globally aggregated models throughout several communication rounds. The selection of clients participating in FL process is currently at complete/quasi randomness. However, the heterogeneity of the client devices within Internet-of-Things environment and their limited communication and computation resources might fail to complete the training task, which may lead to many discarded learning rounds affecting the model accuracy. In this article, we propose FedMCCS, a multicriteria-based approach for client selection in FL. All of the CPU, memory, energy, and time are considered for the clients resources to predict whether they are able to perform the FL task. Particularly, in each round, the number of clients in FedMCCS is maximized to the utmost, while considering each client resources and its capability to successfully train and send the needed updates. The conducted experiments show that FedMCCS outperforms the other approaches by: 1) reducing the number of communication rounds to reach the intended accuracy;2) maximizing the number of clients;3) handling the least number of discarded rounds;and 4) optimizing the network traffic.
In this article, we present a resource allocation and optimization strategy for data center based on resource utilization prediction with back-propagation (BP) neural network, aiming to improve the resource utilizatio...
详细信息
ISBN:
(纸本)9783319700878;9783319700861
In this article, we present a resource allocation and optimization strategy for data center based on resource utilization prediction with back-propagation (BP) neural network, aiming to improve the resourceutilization. We handle resource contention among virtual machines with resource migrating to improve the resourceutilization under the assumption of different functional applications integrated in each server. With the BP network predicted resources utilization and throughput rate of SFC, we adjust and optimize the resource configuration in virtual resource pool and servers, which further improves resourceutilization in data center. Our experiments show that the proposed dynamic resource allocation and optimization strategy performs effectively. And also the BP network achieves more accuracy prediction compared with linear regression model.
暂无评论