Abstract: Random access storage facilities for the CPU of a computer are cascaded in that a facility of relatively fast access speed holds a subset of information held in a facility of lower speed. Memory read requests are applied sequentially to these storage facilities, beginning always with the one of highest access speed, while requests satisfied by a lower speed facilities lead to storage of that information in all facilities of higher access speed. Write requests are made only to one facilities of lower speed, with algorthmic updating of the facility of lowest speed while the ones of higher speed are updated only on subsequent read requests for the same location. The several facilities which can be updated make storage space available on the basis of usages.