A mobile edge computing system with two classes of mobile users is investigated. High priority (subscription) users are allowed more resources at the server than low priority (best-effort) users. Media applications with data encoding are considered so that the encoded data is much smaller than the raw data. Offloading decision is based on task sojourn time which includes communication, computation and queueing delays. Numerical results are presented for optimal service discrimination level between the two classes as well as the effect of queueing delay.