Heterogeneous wireless networks (HWNs) have aroused numerous interest in the past few years in which efficient vertical handoff algorithms are currently attracting much attention. This paper takes into account the vertical handoff decision algorithm in HWNs where multiple types of services with different priorities are supported. By modeling the vertical handoff process as a Markov decision process, a new reward function is introduced which consists of both quality of service profit and handoff cost. Meanwhile in the proposed reward function, multiple network parameters are included for a comprehensive evaluation of the network utilization. Moreover, considering multiple services with different priorities, an online dynamic bandwidth reservation scheme is proposed based on large deviation principle. This scheme, by estimating possible bandwidth request in future time slots, reserves the bandwidth resource …