Due to the exploded demand for mobile traffic and diverse service type, flexible mobile networks with high capacity is drawing lots of research efforts. Distributed large scale (DLS) MIMO has superiority over centralized large scale MIMO and separated small cells due to lower signal loss to users and central control. In this paper, we present a self-optimization method for Cloud RAN based DLS MIMO. The method includes network dimensioning and antenna clustering with adaptability to the migration of hardware/virtualization technology and user over time. Other advantages are easing the process for dimensioning and optimization, requiring fewer antennas, and less demand for controller & baseband computing power due to the characteristic of low complexity.