We need to be consistent and always store sizes in bytes and convert them consistently to human representation when presented to the user. Ref: https://github.com/siemens/mtda/pull/549