Processing large geometric datasets in distributed environments

We describe an innovative Web-based platform to remotely perform complex geometry processing on large triangle meshes. A graphical user interface allows combining available algorithms to build complex pipelines that may also include conditional tasks and loops. The execution is managed by a central engine that delegates the computation to a distributed network of servers and handles the data transmission. The overall amount of data that is flowed through the net is kept within reasonable bounds thanks to an innovative mesh transfer protocol. A novel distributed divide-and-conquer approach enables parallel processing by partitioning the dataset into subparts to be delivered and handled by dedicated servers. Our approach can be used to process an arbitrarily large mesh represented either as a single large file or as a collection of files possibly stored on geographically scattered servers. To prove its effectiveness, we exploited our platform to implement a distributed simplification algorithm which exhibits a significant flexibility, scalability and speed.