One of the promising architectures for the next generation of content distribution networks is the “cache network”, where each node has cache or local storage.We consider the index coding problem, where a central content server disseminates compressed multiple content data to each client that recovers the target content using the cached content, as side information. The coding scheme discussed in this paper is proposed under the condition that there is some correlation between multiple contents, and the scheme is shown to be optimal when the content is less than or equal to three. We believe the sum-product algorithm or linear programming can be used to implement an efficient code by constructing the proposed code using sparse matrices.