Model-As-A-Service (MaaS) Using the Cloud Services Innovation Platform (CSIP)

Abstract

Cloud infrastructures for modelling activities such as data processing, performing environmental simulations, or conducting model calibrations/optimizations provide a cost effective alternative to traditional high performance computing approaches. Cloud - based modelling examples emerged into the m ore formal notion: 'Model - as - a - Service' (MaaS). This paper presents the Cloud Services Innovation Platform (CSIP) as a software framework offering MaaS. It describes both the internal CSIP infrastructure and software architecture that manages cloud resources for typical modelling tasks, and the use of CSIP's ' ModelServices API ' for a modelling application . CSIP's architecture supports fast and resource aware auto - scaling of computational resources. An example model service is presented: the USDA hydrograph model EFH2 used in the desktop - based 'engineering field tools' is deployed as a CSIP service. This and other MaaS CSIP examples benefit from the use of cloud resources to enable straightforward scalable model deployment into cloud environments.