Differentiated Cloud Request Scheduling Strategy And Performance Study

With the advent of the era of big data,cloud computing technology is booming.The key issues of cloud computing are how to satisfy different QoS(Quality of Service)requirements from heterogeneous users and save energy of the cloud data center.In this paper,we research on the differential cloud request scheduling strategies based on the hybrid cloud computing paradigm which the cloud requests can be processed by either local SaaS(Software as a Service)servers or public IaaS(Infrastructure as a Service)cloud.First of all,we divide IaaS cloud data center into rapid cloud and reliable cloud according to the classification of cloud requests with the delay sensitive requests and the fault sensitive requests.Considering the highly reliable service of the fault sensitive requests,we introduce requests retransmission to the reliable cloud,and propose a differential cloud request scheduling strategy based on error recovery.Based on the strategy,considering energy saving of the cloud data center,we introduce dual-rate switch to the VMs in rapid cloud,and propose a differential cloud request scheduling strategy based on rate adjustment.Secondly,according to the differential cloud request scheduling strategy based on error recovery,we establish a hybrid queueing network system model with service feedback.The system performance is evaluated from the view of the average response time of the delay sensitive requests and the utilization of Virtual Machines(VMs)in the reliable cloud.Finally,we provide theoretical experiment and simulation experiment to reveal the tradeoff between different system performances.Thirdly,according to the differential cloud request scheduling strategy based on rate adjustment,we establish a novel hybrid queueing network system model with variable-rate.Then we estimate the system performance by the average response time of the delay sensitive requests,the energy saving level of VMs in rapid cloud and the utilization of VMs in the reliable cloud.Experiments with analysis and simulation are provided to validate the energy saving of the proposed strategy.Finally,for the differential cloud request scheduling strategy based on error recovery,we establish the benefit function of the delay sensitive requests to investigate the Nash equilibrium behavior and the socially optimal behavior.For the differential cloud request scheduling strategy based on rate adjustment,we establish a system cost function according to the tradeoff between different system performances.We improve the Jaya intelligent searching algorithm to formulate a pricing policy for the cloud requests and establish the joint optimization scheme of the VMs' rate D-value and the local SaaS server capacity separately.