Public Cloud Extension for Desktop Applications -- Case Study of a Data Mining Solution

The paper describes challenges and obstacleswhen developing a cloud extension for a computationally expensive desktop application to perform computation tasks in a public cloud. In a case study we highlight thisstep by step with a real-world data mining application and present solutions to realize this scenario. Amazon'sS3, EC2, RDS, SES, IAM and STS services are utilized in a complex setup in order to realize a completely dynamic cloud layer architecture which is needed to implement the extension. This includes the creation of the infrastructure and the task execution in the cloud aswell as its extermination afterwards, which is described besides the data mining application itself. Today's major challenges regarding cloud computing as data security and privacy are taken into account. Additionally, a multicriteria-optimization across different cloud setups is considered in order to guarantee transparency concerning a runtime-cost tradeoff and to rule out suboptimal setups.The approach is based on benchmarks that have to beperformed. The effectiveness of this setup is illustrated by example application instances. The results show, under which circumstances it is beneficial to use the cloud to perform computing tasks.