> you need exact time sync on all nodes for that. no i don't - since i have two timestams (one in the small pkt and another in the big pkt) i can compute the diff and as long as the clock-drift between the two packets is reasonable i can calculate the serialization delay and the bandwidth. /hannes