Research On Cloud Computing Market Forecast And The Combination Of Purchase Strategy For Multi-instances

Cloud computing is a virtualized, scalable IT service, which can dynamically provide services according to the needs of users. With the increasing user demand of the cloud platform, the amount of load resources has been changed a lot. So how to accurately forecast users’ demand and the rational allocation of resources have become face challenges of the cloud platform providers. Thus, a variety of cloud resources market trading platforms has emerged to provide a reliable guarantee for users’timely and convenient access to cloud services.But with the emergence of a variety of cloud computing instances, Such as the Amazon Elastic Compute Cloud platform, it has three types of instances, namely:on-demand instance, reserved instance and spot instance. The first two instances have fix prices, but there are no clear study results about how cloud providers make a price or how the pricing trend goes. In addition, it will affect the efficiency of the final purchase decision that how the users can better obtain the necessary information under cloud computing environments and how they deal with decision-making information from their mass of dynamic demand information. Therefore, studies about the strategy of market transactions and purchase decisions are necessary.The main challenges are as followings:(1) how cloud providers price spot instances and how the prices change are considered only qualitative but quantified now;(2) the own demand information for users cannot be fully processed and analyzed;(3) in the purchase decision-making process, the study nowadays see the quality of service as a only influential factor, unable to meet the various needs of users.Based on the above issues, this paper proposes a cloud-based coalition of cloud computing market trading system for the supply and demand sides of the transaction provides architectural support. Then it builds predictive models of cloud computing market transactions, including forecasting models of spot bid price and customer demands. Finally it studies a combination of multi-instances purchase decisions in order to achieve the goal to meet the diverse needs of users. The main work includes:(1) Cloud computing market trading system.(2) Forecasting mode of cloud computing market spot price. It fully taps the pricing rules of cloud vendor to provide a basis for decision-making tenders;(3)Proposed a demand forecasting model of cloud computing market customer, and uses gray BP neural network simulation and training to makes predictions more accurate and improve the shortcomings of the traditional BP neural network, in order to fully tap the cloud user’s own needs, so that the users can better understand their own needs, then they can make optimized purchasing decisions; (4)Design a portfolio purchase decision model of user-based multi-instance cloud computing market. First considering the smallest cost optimization and service time, combining with the characteristics of different types of instances, it builds two single objective constrained models. Then considering the various needs of customers, it builds a dual objective constraint model, while meeting the customers’ goals of minimum costs and shortest time.