Scalability Analysis of Variational Data Assimilation Algorithms on Hybrid Architectures

Large-scale problems are computationally expensive and their solution requires designing of scalable approaches. Many factors contribute to scalability, including the architecture of the parallel computer and the parallel implementation of the algorithm. However, one important issue is the scalability of the algorithm itself. We have developed a scalable algorithm for solving large scale Data Assimilation problems: starting from a decomposition of the mathematical problems, it uses a partitioning of the solution and a modified regularization functionals. Here we briefly discuss some results.