Applying millimeter wave (mmWave) in mobile edge computing (MEC) networks helps to improve task offloading performance from both communication and computing aspects. In this paper, we propose a MEC-enabled heterogeneous network in which mmWave small cells are deployed overlaid sub-6GHz macro cells. Base stations in two tiers are equipped with MEC servers characterized by different computing capacity and computing buffer. An association strategy based on the MEC service area is designed. In order to analyze the offloading performance, we derive the successful offloading probability and offloading latency with the tools of stochastic geometry and queueing theory. Simulation results demonstrate that there exist optimal values of small base station density and MEC service radius to obtain the maximum successful offloading probability and the minimum offloading latency, which provides useful guidance for the practical deployment.