Millimeter wave (mmWave) communication is apromising candidate for future extremely high data rate, wirelessnetworks. The main challenges of mmWave communications aredeafness (misalignment between the beams of the transmitterand receiver) and blockage (severe attenuation due to obstacles).Due to deafness, prior to link establishment between a clientand its access point, a time consuming alignment/beam trainingprocedure is necessary, whose complexity depends on the operatingbeamwidth. Addressing blockage may require a reassociationto non-blocked access points, which in turn imposes additionalalignment overhead. This paper introduces a unifying frameworkto maximize network throughput considering both deafness andblockage. A distributed auction-based solution is proposed, wherethe clients and access points act asynchronously to achieveoptimal association along with the optimal operating beamwidth.It is shown that the proposed algorithm provably converges toa solution that maximizes the aggregate network utility withina desired bound. Convergence time and performance boundsare derived in closed-forms. Numerical results confirm superiorthroughput performance of the proposed solution compared toexisting approaches, and highlight the existence of a tradeoffbetween alignment overhead and achievable throughput thataffects the optimal association.