Abstract:
|
The Dynamic Pipelineis a concurrent programming pattern amenable to be parallelized. Furthermore, the number of processing units used in the parallelization is adjusted to the size of the problem, and each processing unit uses a reduced memory footprint. Contrary to other approaches, the Dynamic Pipeline can be seen as ageneralization of the (parallel) Divide and Conquer schema, where systems can be reconfigured depending on the particular instance of the problem to be solved. We claim that the Dynamic Pipelines is useful to deal with Big Data related problems. In particular, we have designed and implemented algorithms for computing graphs parameters as number of triangles, connected components, and maximal cliques, among others. Currently, we are focused on designing and implementing an efficient algorithm to evaluate conjunctive query. |