This paper addresses the question of maximizing throughput in multi-hop wireless networks given a specific placement of nodes and a specific traffic workload. Unlike previous work that focuses on asymptotic performance bounds under assumptions of homogeneity or randomness, this paper works with any given network and workload specified as inputs. A key issue is wireless interference between neighboring nodes, which is modeled using a conflict graph. The authors present methods for computing upper and lower bounds on the optimal throughput for the given network and workload. They assume that packet transmissions can be finely controlled and scheduled by an omniscient central entity, which is unrealistic but provides a best-case bound. Using ns-2 simulations, they show that routes derived from their analysis often yield significantly better throughput than default shortest path routes, even in the presence of uncoordinated packet transmissions and MAC contention. This suggests that achieving throughput gains is possible by employing an interference-aware routing protocol. The paper also evaluates how per-node throughput varies as the number of nodes grows, finding that the addition of new nodes can improve per-node throughput due to increased connectivity and routing around interference hotspots. The authors discuss the limitations of their work and conclude by highlighting the generality of their methodology and the conflict graph framework.This paper addresses the question of maximizing throughput in multi-hop wireless networks given a specific placement of nodes and a specific traffic workload. Unlike previous work that focuses on asymptotic performance bounds under assumptions of homogeneity or randomness, this paper works with any given network and workload specified as inputs. A key issue is wireless interference between neighboring nodes, which is modeled using a conflict graph. The authors present methods for computing upper and lower bounds on the optimal throughput for the given network and workload. They assume that packet transmissions can be finely controlled and scheduled by an omniscient central entity, which is unrealistic but provides a best-case bound. Using ns-2 simulations, they show that routes derived from their analysis often yield significantly better throughput than default shortest path routes, even in the presence of uncoordinated packet transmissions and MAC contention. This suggests that achieving throughput gains is possible by employing an interference-aware routing protocol. The paper also evaluates how per-node throughput varies as the number of nodes grows, finding that the addition of new nodes can improve per-node throughput due to increased connectivity and routing around interference hotspots. The authors discuss the limitations of their work and conclude by highlighting the generality of their methodology and the conflict graph framework.