OGSTM - BFM I/O IMPROVEMENTS

This thesis project is part of a restyling work on the OGSTM - BFM model that is used to study the biogeochemical properties of Mediterranean Sea waters. OGS is one of the research centres that are involved in the Copernicus (European Union's Earth observation programme) project, being part of the 6th MSC (Monitoring and Forecasting Centres) which aim is to provide regular and systematic information about the physical state of the ocean and marine ecosystems for the Mediterranean Sea. OGS runs the OGSTM - BFM model (ocean numerical model) assimilating TAC (Thematic Data Assembly Centres) data to generate reanalyse (20 years in the past), analyse (today) and 10-days forecasts of the Mediterranean Sea. This thesis is aimed at the improvements of the outputs to enhance the dumping capability without affecting the total run time. Before this thesis, the output of the model was serialised on a single processor. This way of handling I/O was time consuming as the new Copernicus project requests asked to save daily the model output (before was saved in part weekly and in part monthly). The first aim of the thesis was to generate a distributed parallel I/O in which there are more writing processes that in parallel dump different output variables, in a way to scale down writing time. After the implementation of this paradigm the focus was directed to apply the same paradigm to the assimilation part (3DVAR) of model, as the daily assimilation had to be introduced (before was weekly).