Novel energy-aware approach to resource allocation in cloud computing

Abstract

In this paper, we address the issue of resource allocation in a Cloud Computing environment. Since the need for cloud resources has led to the rapid growth of data centers and the waste of idle resources, high-power consumption has emerged. Therefore, we develop an approach that reduces energy consumption. Decision-making for adequate tasks and virtual machines (VMs) with their consolidation minimizes this latter. The aim of the proposed approach is energy efficiency. It consists of two processes; the first one allows the mapping of user tasks to VMs. Whereas, the second process consists of mapping virtual machines to the best location (physical machines). This paper focuses on this latter to develop a model by using a deep neural network and the ELECTRE methods supported by the K-nearest neighbor classifier. The experiments show that our model can produce promising results compared to other works of literature. This model also presents good scalability to improve the learning, allowing, thus, to achieve our objectives.

Fault Tolerance for Composite Cloud Services: A MAS-based Novel Approach

“…To deal with this problem, we select another path by using the approach proposed in section 3. The compositions services plans, for the composites services CS [2], CS [4] and CS [5], are constructed as the same manner as the composition plan of the CS [3]. Table 6 summarizes the different composition plans for the CS [i] where i [2..5].…”

“…The objective of this scenario is to show the impact of alternative services number on the total cost. In this experimentation, for each composite service CS [2], CS [3], CS [4] and CS [5], we vary alternative services number from 3 to 10, and we observe the total cost for each CSi. In this simulation, we do not deal with the unavailable services.…”

“…It is designed to run large-scale applications. This is a cost-effective way to achieve the Service Level Agreement (SLA) and Quality of Service (QoS) according to application demand [2].…”