Abstract-Current approaches for designing wireless mesh networks (WMNs) suffer from low-end-to-end-throughput performance. This disadvantage mainly comes from the traffic aggregation and multihop properties of WMNs. Although the throughput performance can be improved by utilizing multiple channels and multiple radios, the knowledge of how good the performance can be is still unavailable. Some previous works endeavored to heuristically improve the systems' performance, while others focused on the optimization algorithms for satisfying the clients' traffic demand. There is no work, until now, that can provide the knowledge of the system capacity. Although optimization algorithms have been proposed in previous papers, they require the exact traffic load input and make the "optimal" result easily outdated due to local topology and traffic changes. In this paper, we investigate the capability that WMNs can offer the mesh clients to deliver/receive data across the gateways, which is termed the portal capacity. Furthermore, we propose our solution to achieving the optimality in portal capacity through the centralized algorithm while retaining optimality via distributed tuning when local changes in either topology or traffic occur in the system. The concept of portal capacity provides us with very useful information about a WMN's capability, which can facilitate the optimization of the end-to-end throughput and fairness. In addition to this, this paper is the first one that takes the optimality and dynamic property of the system into consideration.Index Terms-Channel assignment, multiple channels and multiple radios (MC-MR), resource allocation, wireless mesh network (WMN).