Random Task-oriented User Utility Optimization Model In The Cloud Environment

In order to meet the advent of the information age, many enterprises andorganizations deploy their private information systems. They need to purchasecomputer hardware and software to build their information systems, and organize aspecial department as well as technicians for the deployment, management andmaintenance of the systems. Holding a private information system for each enterpriseor organization consumes a large amount of computer resources and it is very hard tomanage. In addition, it is needed to purchase more computer resources to expand thescale of the information system when necessary. And if the systems are over-deployed,a great amount of computer resources are wasted.Cloud computing can solve above problems fundamentally. It virtualizes variousresources, including servers, storage devices, and network equipments, as a unifiedresource pool by virtualization technology. The implementation detail of theunderlying resources is hided for users. The users can obtain hardware, software andplatform resources through network. Cloud computing is a large heterogeneoussystem, which consists of various systems and applications. In order to satisfy theusers’ requirement, save computer resource, and optimize the system performance, itis necessary to allocate the resources reasonably. Therefore, the resource allocatinghas become one of the hotspots in the field of cloud computing.On the basis of research results at home and abroad, this paper firstly summarizesthe research status of cloud computing and some related preliminaries, and then therelated models and scheduling algorithms in cloud computing are introduced, such asstochastic system model. At last, according to the randomness of the tasks in the cloudenvironment, this paper proposes a novel task scheduling model as well as itsoptimization algorithm. A lot of experiments are done to validate the efficiency of ourproposed model. The main contributions of this paper are listed as follows. First,according to the randomness of user tasks requiring the cloud resources, this paperintroduces stochastic systems theory (queuing theory) to build the model of user tasksaccessing. Second, it introduces the concept of the user utility in economics,proposing the user utility description model in the cloud environment. Third,according to the different performance requirement, it divides the users into twocategories, saying time-sensitive and cost-sensitive, and two corresponding utility functions are given. Fourth, the formal description of user utility is given for twomodels respectively. It regards the maximum user utility values as a goal. Besides, itbecomes constrained by some objective factors, such as the waiting time in the queue,execution time, cost, parallel speedup ratio of the user task. Finally, it realizesdynamic allocation of the computing resource, and offers the optimization algorithmof user utility model. The algorithm takes the user utility as a standard to optimize thequeued sequential of the user tasks, aiming at maximizing the user utility of each user.At last, the experiments show that the optimization algorithm can satisfy users’demand in terms of execution time and cost, and maximize the total utility of user atthe same time.