We study opportunistic scheduling in a multiuser wireless network where there are K users, each with a backlogged message for transmission. A key goal is to minimize the total completion time via a cross-layer design approach. Our main contribution consists of two size-aided opportunistic scheduling (SAOS) schemes. The SAOS schemes use both the traffic information and channel variation to reduce the completion time. Our results show that the SAOS schemes can yield significant reduction in the total completion time. We also examine two extreme cases, i.e., wireless shortest processing time first (SRPT) and "riding on the channel peak", which utilize only either the file size information or the instantaneous channel conditions.