A Cloud Environment for Ubiquitous Medical Image Reconstruction

The new needs that have appeared in recent years around the medical image reconstruction field have pushed a greater investment on computing and storage platforms in order to keep up with the increased requirements in performance and quality. To reduce the cost of purchasing and managing these resources novel ubiquituos services have arisen. In this work we describe and evaluate a cloud environment for ubiquitous medical image reconstruction based on Amazon EC2 and a heterogeneous Apache Spark platform. We show an architecture composed by a client application, a front-end server and an isolated cluster per application for the uninterrupted execution of the reconstruction service. We present an execution and cost model for the prediction and configuration of the cloud environment based on performance and monetary costs. The evaluation shows the best configurations in terms of execution time for different types of instances and tuning parameters. Based on these results we compute a Pareto front to study the best solutions for the configuration of the cloud environment with the objective of minimizing both execution time and costs.