On the successful coexistence of oceanographic operational services with other computational workloads

Abstract

This paper describes a case of the implementation of a pan-European operational oceanography service coexisting with non-operative research jobs that are executed on the same computational resources. The complexity of designing a good operational service increases when the resources that will be used are shared with other computational workloads. However, once the implementation is achieved the result is an optimised and robust service. Computational resources and other necessary services are permanently monitored in order to detect and solve potential problems in real-time. These resources can be used by other researchers during the time interval in which they are not needed for this operational service. The goal of this work is to guarantee the time to solution of the operational execution on a shared computing environment without a large impact on the researchers jobs.