A computer-implemented method for layered storage of enterprise data comprises receiving from one or more virtual machines data blocks; de-duplicating the data blocks per hypervisor; storing de-duplicated data blocks in a local cache memory; time-based grouping the data blocks into data containers; dividing each data container in X fixed length mega-blocks; for each data container applying erasure encoding to the X fixed length mega-blocks to thereby generate Y fixed length mega-blocks with redundant data, Y being larger than X; and distributed storing the Y fixed length mega-blocks across multiple backend storage systems.