The Research And Implement Of Elastic Decision Mechanism In Cloud

As a rising computer science direction, cloud computing owns a huge potential in thecurrent social environment. The users obtain services and resources from the computingplatform and pay as they use, which make it possible for the uses to satisfied themselves atreasonable cost.The most direct reason for the cloud computing to raise widespread concern is theeconomic factor. Users can avoid providing infrastructure and maintenance for theirapplications through renting the infrastructure and other resources in the cloud. In addition,the cloud platform owns the feature of providing resources as users command, whichmakes it possible for the users to control their applications further. All these things makethe cloud computing the first choice of the most people.Many of the commercial cloud platforms can provide the ability of elastic support,such as Amazon EC2and Windows Azure and so on. For example, the Auto Scalingservice in Amazon EC2can support the ability of scaling out and scaling in on the numbersof the EC2service numbers when working with the CloudWatch Service. Users can definetheir own rules to scale out and scale in through creating the trigger. But this kind ofcontrol approach is something too rigid to make the decision. All in all, this kind of triggercan provide the elastic support for the users, but can also bring the unnecessary cost to theusers.An appropriate approach to manage computing resources is necessary if the usesexpect the high performance and the low cost at the same time. Usually, the manual controlis the most choice to solve this problem. While the applications in the cloud being moreand more complex, the manual way is becoming unacceptable. Thus, it is a meaningful andchallenging task to establish a elastic decision mechanism according to whose strategiesthe applications can performance well at the low cost.To deal with this problem, an elastic decision mechanism is proposed in this paper.This elastic decision mechanism will establish the system workload model and the systemperformance model when it starts to work, and make the elastic decision according to theoutput of these two models. Meanwhile, the elastic decision mechanism will checkwhether these two models are effect or not while managing the computing resources of the users’ applications to prevent these models from becoming invalid. And it will update orrebuild the models according to the results of the check.At the last of this paper, an experiment will be proceeded to verify the function andthe performance of the elastic decision mechanism. The elastic decision mechanism willdemonstrate the ability of elastic support in the experiment. And the actions andcharacteristics of it will also be demonstrated.