Counting triangles (i.e., cliques of size three) is a classical graph problem with many applications, including anomaly detection, community detection, and link recommendation. For triangle counting in large and dynamic graphs, recent work has focused largely on streaming algorithms and distributed algorithms. Can we have the best of both worlds? The objective of this project is to develop a distributed streaming algorithm that utilizes multiple machines to estimate the count of triangles in a graph stream. Specifically, we aim to develop a distributed streaming algorithm that is (a) Fast and accurate: should estimate the number of triangles in the input graph stream significantly faster and more accurate than existing streaming algorithms, (b) Scalable: should run in linear time, and (c) Theoretically sound: should give unbiased estimates whose variances decrease rapidly as the number of machines is scaled up. Alongside, we plan to extend our algorithm to fully dynamic graph streams where both insertions and deletions of edges are streamed over time. Lastly, our algorithm will be a part of our distributed graph-stream processing system, which we plan to develop as a long-term future work.