A Model-driven Approach for Price/Performance Tradeoffs in Cloud-based MapReduce Application Deployment.
Shashank Shekhar, Faruk Caglar, Kyoungho An, Takayuki Kuroda, Aniruddha S. Gokhale, Swapna S. Gokhale · 2013
This paper describes preliminary work in developing a modeldriven approach to conducting price/performance tradeo s for Cloudbased MapReduce application deployment. The need for this work stems from the signi cant variability in both the MapReduce application characteristics and price/performance characteristics of the underlying cloud platform. Our approach involves a model-based machine learning capability that trains itself from executing a variety of MapReduce applications on di erent cloud service providers, and in turn providing useful price/performance tradeo information to MapReduce application users. Additionally, the model-based platform serves to automate the deployment of a MapReduce application to the cloud by incorporating the tradeo choices.