A data storage device includes a controller connected via a plurality of channels to a plurality of clusters, wherein each cluster comprises a scale-out device including a scale-out controller and a buffer. The scale-out controller is connected to a plurality of sub-channels, each one of the plurality of sub-channels connecting a group of non-volatile memory (NVM) devices, such that the scale-out controller controls execution of data processing operations directed to any one of the NVM devices and the buffer.