The Study And Realization On Elasticity Evaluation Model In Cloud Computing

Cloud computing is a kind of model that enables user ubiquitous, convenient,on-demand access to resources that can be provisioned and released quickly from a shared resource pool through the Internet. Elasticity is a very important feature of cloud computing, which make it possible for dynamic allocation of resources. In a cloud environment, when the load rising and the services provided by the providers has violated the SLA(Service Level Agreement), Service providers need to increase the resources to maintain the quality of service. if the load down, they need to reduce the resources to avoid waste.Users need to measure the performance of different cloud services like the traditional services before be used, elasticity is one of the important metrics that should be considered. Elasticity has caused more and more attention now, so the elastic evaluation is becoming more and more important. Now there are a lot of research on elastic evaluation methods, including some benchmark and some methods from the perspective of resource allocation precision and cost. The allocation of resources is calculated mainly through the difference between the number of required resources and the number of actual provisioned resources. And cost calculation, mainly based on SLA, use the method of calculating penalty, to calculate the summation the penalty for over-provisioning and the penalty for underprovisioning.Elasticity’s core features are the speed and precision of the resources allocation, the existing evaluation methods only consider speed or they only consider precision, and only from one perspective of user or provider to measure e lasticity. So lack of a comprehensive method for measuring elasticity. To solve the above problem, this paper made a thorough research for the elasticity characteristics of cloud services, as well as the existing elasticity evaluation methods. Studied and analyzed the characteristic of elastic cloud services. Make analysis on the factors for a complete evaluation system. A complete evaluation system should first determine the different scaling strategies for different cloud platform, and then design elastic ity evaluation metrics conform to the requirement of people, finally design different kinds of workloads to test the cloud platform. This article conducted a more comprehensive analysis from the perspectives of resource allocation, QoS and the resource al location time. Proposed an evaluation method applied to both providers and users. Moreover, designed computational models for resource allocation and the resource allocation time, in addition to improving existing penalty model. Finally, TPC-W benchmarks were tested on Cloud Stack platform with two elasticity scaling strategies namely auto-scaling and scale-out to verify the method.