The role of feedback in rewritable storage channels [Lecture Notes]

New memory technologies and the intrinsic issues associated with device scaling are making the behavior of storage systems increasingly similar to that of noisy communication channels. The ability to read and possibly rewrite content onto memory, however, makes rewritable channels fundamentally different from standard point-to-point channels from an information theoretic standpoint. In this article, we have formalized the problem of finding the storage capacity of a rewritable medium. For a simple uniformly distributed noise model, we derived upper and lower bounds on the storage capacity. An extension to data-dependent noise was also presented. These results provide further evidence of the value of rewriting not only as a pedagogical tool but also as an essential strategy in making nonvolatile memories market competitive.