The emergence of Internet-of-Things (IoT) aims to improve modern life by inter-connecting many smart devices, technologies, and applications. It, however, drastically increases mobile traffic and requires large wireless spectrum. Therefore, for a better spectrum utilization to support this massive connectivity along with low latency and close proximity based communication, non-orthogonal multiple access enabled underlay device-to-device (D2D-NOMA) is envisioned as a key technology for IoT deployment. Hence, in this work, we consider multiple IoT clusters communicating via underlay D2D-NOMA in an uplink cellular network, with the objective of minimizing the outage probability (OP) while allocating resources to the clusters. Specifically, for each cluster, we formulate the optimization problem of OP-based power control with fixed channel allocation (PCCA) and solve it using particle swarm optimization (PSO). Additionally, we prove the convexity of PCCA problem and obtain closed-form expressions at high SNR. Further, we propose a novel OP-aware channel allocation (OPCA) algorithm to enhance the overall D2D network outage performance. The results are numerically validated and an average performance improvement of 31% and 51% is observed over the two conventional schemes namely fixed power allocation and orthogonal multiple access enabled D2D, respectively.