k-means problem is one of the important formulations of the clustering. The k-means objective function nicely 
models the property that, similar (or closely located points) should be in the same cluster and dissimilar 
(or far-away points) should be in different clusters. We sometime want the clusters to follows some additional 
constraints such as: each cluster should roughly contain same number of points. This gives rise to the constrained
 versions of the k-means problem. We will show that there is a 4-pass, logspace streaming PTAS(for fixed k) for 
various constrained versions of the k-means problem