Batch codes, introduced by Ishai et al. in [11], are methods for solving the following data storage problem: n data items are to be stored in m servers in such a way that any k of the n items can be retrieved by reading at most t items from each server, and that the total number of items stored in m servers is N. A combinatorial batch code (CBC) is a batch code where each data item is stored without change, i.e., each stored data item is a copy of one of the n data items. One of the basic yet challenging problems is to find optimal CBCs, i.e., CBCs for which total storage (N) is minimal for given values of n, m, k, and t. In [13], Paterson et al. exclusively studied CBCs and gave constructions of some optimal CBCs. In this article, we give a lower bound on the total storage (N) for CBCs. We give explicit construction of optimal CBCs for a range of values of n. For a different range of values of n, we give explicit construction of optimal and almost optimal CBCs. Our results partly settle an open problem of [13]. © 2012 AIMS.