Prediction model for suitability of resource deployment using complex templates in OpenStack

Enterprises are moving into cloud, so deployment and configuration of virtual machines and application software are becoming highly complex. The solution to this problem is to automate the deployment and configuration using templates. These templates allow the users to specify the information about resource requirements of an application. In this paper, firstly, we present the methods for development of complex templates that consist of various properties like input parameters and conditional logic, lacking in existing simple templates. OpenStack is the popularly used private Cloud deployment tool and many other similar tools lack in providing the complex templates in their architecture. The complex templates proposed in this paper are generic in nature and designed keeping OpenStack in mind. Using the proposed complex templates, based on user defined input values and conditional logic, various infrastructure stacks can be created in simple templates and can be directly fed into heat engine for deployment. The complex templates can be developed to accept input parameters, including conditional logic or auto scaling constructs. Though the complex template allows us to specify the requirement in an easy way but it is difficult and sometimes inaccurate, to visualize the outcome. Under such circumstances estimating the billing cost of the infrastructure stack is extremely difficult. In order to alleviate this problem, in this paper secondly, we propose a system that can automate the cost estimation of a stack before its creation. In order to assess the cost estimates, we have implemented machine learning using decision trees on the templates data that can produce the output in the form of two classes and give the decision that whether it is good to deploy templates or not. The preliminary assessment is done on the proposed model and results achieved are produced.