Massive multiple-input-multiple-output (MIMO) base station deployments and millimeter wave (mmWave) spectrum utilization have been identified as promising disruptive technologies, along with ultra-dense Heterogeneous Networks (UDHNs) to meet the exponential data requirement of the next generation cellular networks. With the proliferation of UDHNs, optimal user-cell association in cellular networks, which is a well-known open problem, will be exacerbated due to the power differential of macro and small cells. This study investigates the user-cell association problem for ultra-dense two-tier networks with massive MIMO deployment and small cells operating in mmWave spectrum. The association problem is modeled as a convex utility maximization problem, adapted from [11], and is a function of the user throughput. The problem is solved through a centralized subgradient algorithm. Additionally, a game theoretical user-centric distributed load balancing algorithm, inspired from [32], where each user chooses its serving base station to maximize its user throughput selfishly, is also evaluated. Moreover, these adapted algorithms are compared against smallest pathloss and maximum downlink data rate association methods and it is demonstrated via extensive simulations that both the centralized and user-centric approaches almost equally outperform the smallest pathloss and maximum downlink data rate association methodologies in terms of user throughput and cell load distribution. The results exhibit average throughput gains between 20% and 40% for the majority of users if massive MIMO UDHN deployments are operated in the mmWave spectrum as compared to existing sub-6 GHz bands under the optimal user-cell association schemes.