Techniques are described for load balancing subscriber sessions across tunnel termination devices. A network device is described, for example, that includes a tunneling module that load balances subscriber sessions across a plurality of tunnel termination devices based on weightings associated with the tunnel termination devices. The weightings may be assigned to the tunnel termination devices by a user, or may be calculated by the network device based on resource constraints associated with the tunnel termination devices. The network device may calculate the weightings, for example, based on a maximum number of subscriber sessions supported by each of the tunnel termination devices. As one example, the techniques may be applied to load balance Point-to-Point (PPP) subscriber sessions across L2TP Network Servers (LNSs).