Do we need to do feature scaling on each dimension of a multi-dimensional data distribution before applying K-Means clustering algorithm for it to be effective ?
Yes, in general, attribute scaling is important to be applied with K-means. Most of the time, the standard Euclidean distance is used (as a distance function of K-means) with the assumption that the attributes are normalized.
Yes, to make sure that your calculations will not be biased either to the very high or to the very low values. In other words, to make sure that all your data are at the same level. you could use any normalization technique to do this, and I recommend this:
Xi(new)= (Xi-mean(all X values))/standard deviation
In most cases yes. But the answer is mainly based on the similarity/dissimilarity function you used in k-means. If the similarity measurement will not be influenced by the scale of your attributes, it is not necessary to do the scaling job.