Exploiting parallelism to support scalable hierarchical clustering