Today, mobile consumers increasingly use Augmented Reality (AR) devices to stream personal video through carrier networks. Thanks to its flexibility, Software-Defined Networking (SDN) is deployed in many carrier networks to support end-to-end network-slicing, which is substantial for these AR applications. In an OpenFlow-enabled SDN network, a controller must decide the rules to be placed into the switches in the network, subject to multiple constraints such as memory capacity, link bandwidth limitation, and flow continuity. Due to legacy switch models, prior work focuses only on unicast flows which cannot efficiently support AR applications with streaming traffic from one source device (or server) to many destination devices across the network. In this paper, we optimize rule placement in resource-constrained Openflow networks for both unicast and multicast flows. Our approach is to leverage the use of Group Tables, which is recently introduced in the OpenFlow 1.1 specification, to support multicast flows and, at the same time, save switch memory. Traffic to multiple destinations can be aggregated to match a single flow table entry per switch. Therefore, significant link resources can be saved. The experimental results on three different topologies show our solution can support a higher number of flows than the state-of-the-art solutions by reducing both the link usage by up to 30% and the number of flow entries needed to deliver the traffic to destinations by 22%.