Mathematical optimization techniques often require immense computational power to perform in a timely manner, which we historically have not had available. Though there have certainly been significant advances in optimization techniques, the bottleneck of solving large problems is still the scarcity of computational resources. With the advent of large-scale cloud computing infrastructures, we can leverage them to finally solve large optimization problems. The main issue is that we need not only an optimization framework, but a cloud framework that can properly utilize the immense quantity of resources that are necessary to run the optimization framework. Project Saturn aims to develop such a generic cloud-based optimization framework.
The Saturn framework is designed to provide a complete library of single- and multi-objective global optimization algorithms, which can be utilized both on local computing resources as well as on the cloud. Projects can integrate Saturn directly as a multi-language multi-platform optimization library with a dedicated API or through a web-service based API to communicate with the global optimization framework running inside the Autodesk Cloud. Results can be seen through Saturn's own advanced visualization toolkit, which is available through its web interface.
The ultimate goal is to empower people with an easily-accessible optimization framework that can leverage the immense computational power available in the cloud to obtain the results they are seeking. Saturn is therefore modular and customizable - users can tune and modify existing optimization algorithms as well as design their own optimization algorithms and objective functions to satisfy their individual needs, and freely compose them into complex large-scale optimization processes. Algorithms don't even need to be optimization tasks, as Saturn is designed to be a system for performing general-purpose massively parallel operations.By simply creating an algorithm that creates tasks Saturn will distribute the tasks locally and into the cloud by efficiently utilizing all the computational resources available.