Transferable Knowledge for Low-Cost Decision Making in Cloud Environments

Users of Infrastructure as a Service (IaaS) are increasingly overwhelmed with the wide range providers and services offered by each provider. As such, many users select based on description alone. An emerging alternative is to use decision support system (DSS), which typically relies gaining insights from observational data in order assist customer making decisions regarding optimal deployment cloud applications. The primary activity such systems generation prediction model (e.g. using machine learning), requires significantly large amount training data. However, considering varying architectures applications, providers, offerings, this not sustainable it incurs additional time cost collect train models. We overcome through developing Transfer Learning (TL) approach where knowledge (in form associated set) gained running an application particular IaaS transferred substantially reduce overhead building new models for performance applications and/or infrastructures. In article, we present our evaluate extensive experimentation involving three real world over two major public namely Amazon Google. Our evaluation shows that novel two-mode TL scheme increases overall efficiency factor 60 percent reduction generating model. test under number cross-application cross-cloud scenarios.