Novel solid state drive storage technologies promise  faster data access  and
higher energy efficiency than conventional hard disk drives. To fully utilize
these emerging technologies, it is necessary to develop error-correction
schemes
attuned to the physical properties of the memory devices. Based on
experimentally-collected data from a triple level cell (TLC) Flash drive, we
propose a novel error correction model that exploits the underlying asymmetry
in
device variability so that only certain intra cell level changes are
permitted.
New error correction codes build upon classical Tensor product ideas and
demonstrate a substantial lifetime improvement over conventional schemes
routinely designed for symmetric error correction.