Impact of Reverse Traffic and Node Mobility on Dack Performance in Manets

TCP performs poorly in mobile ad hoc networks. Previous work showed significant TCP goodput improvements resulting from the use of the delayed TCP ACK option (DACK). In this paper, a simulation study of the impact of DACK on TCP goodput in mobile ad hoc networks is presented. Two key typical assumptions have been removed: only forward-direction data traffic is present and nodes are static. It is shown that, in the presence of even a moderate amount of reverse data traffic, DACK does not necessarily always result in TCP goodput improvements; and when it does, the improvements are less than what was reported previously. The sensitivity of this finding to the TCP variant used and type of reverse data traffic (whether UDP or TCP), have been investigated, and it is shown to continue to be valid for most scenarios examined. It is also shown that node mobility reduces the benefits of DACK to the point of harming TCP goodput, especially with large number of nodes