Hierarchical Clustering of Massive, High Dimensional Data Sets by Exploiting Ultrametric Embedding