As migration to the cloud becomes more ubiquitous, application owners need to determine the feasibility and, most critically, cost of migration. Currently, $30-75K of in-depth analysis is typically required to estimate the cost of migration for each application, causing a barrier to entry for some applications and creating a backlog in the hundreds of applications considering migration. The Cloud One Application Migration CER is a less resource-intensive estimating model that can provide an accurate estimate and facilitate buy down of the backlog.

The Cloud One Application Migration CER is a multiple regression model that predicts the costs of application migrations from legacy DOD environments to commercial cloud hosting environments. Independent variables were selected from a group of software application technical characteristics using backwards stepwise selection. Model validation was performed using Leave One Out Cross-Validation (LOOCV) and models were tested using the Predicted Residual Error Sum of Squares (PRESS) metric. Cost data was provided by the Cloud One program office, extracted from financial reports provided by the contractor system integrator. Technical data was extracted from application analysis reports provided prior to application migration.

The focus of the talk will be the methodologies for and results of a four year-long data collection and analysis effort on the cost and technical drivers of migrating legacy Air Force applications to the Air Force’s commercial cloud environments in Cloud One. This should be a great session for anyone interested in CER development, linear models, data collection and preparation, or cloud computing and application migration.