Document Type
Presentation
Publication Date
8-1-2014
Abstract
The traditional paradigm for the deployment of hydrological models involves the capturing and testing of model concepts and numerical consistency for robustness and accuracy, which is then distributed as binary files with or without source code. The model software is then populated with data and parameters and run locally within the modeller’s organisation, often on their own desktop. This modelling workflow is used by many organisations; however, there are several limitations and potential issues. Once the software is outside the developer’s organisation they rely on the modeller to apply updates and bug fixes in a timely manner, and to correctly describe the model version used for reporting. The developer also loses control of the quality and suitability of the input data for a particular application of the model. With more prevalent access to high bandwidth internet and flexible computing infrastructure there is an increased opportunity to better control model access through the exposure of modelling functionality through web services. As well as giving the developer tighter control over model versioning and IP, it also allows closer coupling of the model to both data sources and computational resources, which is especially beneficial to multi-run use cases such as uncertainty analysis and calibration, where the ability to easily scale to many model instances is of most value. The eWater Source modelling system is an important use case for Australia’s hydrologic community, and provides a rich array of functionality. Source is especially suited to the services modelling paradigm as it has project load times much greater than simulation runtimes, the services based approach allows the hiding of these load times by keeping the project in memory for each instance of a Source Server. This paper investigates the use of a Source service interface for providing hydrological modelling web services.
Comments
Session R07, Web Services and Cloud Computing