Clustering problems are well known in database literature for their use in numerous applications. Multidimensional data always is a challenge for clustering algorithms. The Halite, fast and scalable clustering method that looks for clusters in subspaces of multidimensional data. The tree root corresponds to a hypercube embodying the full data set. The next level divides the space in a set of 2D hypercube. The resulting hypercube are divided again, generating the tree structure. Bump Hunting task refers to apply for each level of the Counting-tree one d-dimensional Laplacian mask over the respective grid to spot bumps in the respective resolution. Specifically the main contributions of Halite are: Scalability: it is linear in time and space regarding the data size and dimensionality of the clusters’ subspaces. Usability: it is deterministic, robust to noise, doesn’t take the number of clusters as an input parameter, and detects clusters in subspaces generated by original axes or by their linear combinations, including space rotation. Effectiveness: it is accurate, providing results with equal or better quality. It is achieved through word based approach Generality: it includes a soft clustering approach.