Maintaining Data Availability (DA) is a big challenge in Decentralized Online Social Networks (DOSN). Nowadays, the limitation of the storage capacity in DOSN becomes a critical factor that jeopardizes the DA. Therefore, it is desired to determine the relation between the storage capacity of DOSN and the level of DA, and develop an approach to mitigating the limitation of storage capacity. This paper addresses these issues. In this paper, a probabilistic DA model over storage capacity is established. A novel method is then proposed to predict the DA on the fly. Further, a Cloud-assisted DOSN (CDOSN) framework is proposed to enhance the storage capacity and the DA in DOSN. This paper conducts the detailed quantitative analysis about the storage capacity and the DA in CDOSN. Extensive simulation experiments have been conducted to evaluate the effectiveness of the DA model, the on-the-fly prediction and the CDOSN framework.