Banking theory based distributed resource management and scheduling for hybrid cloud computing

Cloud computing is a computing model in which the network offers a dynamically scalable service based on virtualized resources. The resources in the cloud environment are heterogeneous and geographically distributed. The user does not need to know how to manage those who support the cloud computing infrastructure. From the view of cloud computing, all hardware, software and networks are resources. All of the resources are dynamically scalable on demand. It can offer a complete service for the user even when these service resources are geographically distributed. The user pays for only what they use (pay-per-use). Meanwhile, the transaction environment will decide how to manage resource usage and cost, because all of the transactions have to follow the rule of the market. How to manage and schedule resources effectively becomes a very important part of cloud computing, and how to setup a new framework to offer a reliable, safe and executable service are very important issues.

The approach herein is a new contribution to cloud computing. It not only proposes a hybrid cloud computing model based on banking theory to manage transactions among all participants in the hybrid cloud computing environment, but also proposes a "Cloud Bank" framework to support all the related issues. There are some of technology and theory been used to offer contributions as below:

1. This thesis presents an Optimal Deposit-loan Ratio Theory to adjust the pricing between the resource provider and resource consumer to realize both benefit maximization and cloud service optimization for all participants.

2. It also offers a new pricing schema using Centralized Synchronous Algorithm and

Distributed Price Adjustment Algorithm to control all lifecycles and dynamically price all resources.

3. Normally, commercial banks apply four factors mitigation and to predict the risk:

Probability of Default, Loss Given Default, Exposure at Default and Maturity. This thesis applies Probability of Default model of credit risk to forecast the safety supply of the resource. The Logistic Regression Model been used to control some factors in resource allocation. At the same time, the thesis uses Multivariate Statistical analysis to predict risk.

4. The Cloud Bank model applies an improved Pareto Optimality Algorithm to build its own scheduling system.

5. In order to archive the above purpose, this thesis proposes a new QoS-based

SLA-CBSAL to describe all the physical resource and the processing of thread.

In order to support all the related algorithms and theories, the thesis uses the CloudSim simulation tools give a test result to support some of the Cloud Bank management strategies and algorithms. The experiment shows us that the Cloud Bank Model is a new possible solution for hybrid cloud computing.

For future research direction, the author will focus on building real hybrid cloud computing and simulate actual user behaviour in a real environment, and continue to modify and improve the feasibility and effectiveness of the project. For the risk mitigation and prediction, the risks can be divided into the four categories: credit risk, liquidity risk, operational risk, and other risks. Although this thesis uses credit risk and liquidity risk research, in a real trading environment operational risks and other risks exist. Only through improvements to the designation of all risk types of analysis and strategy can our Cloud Bank be considered relatively complete.