Much of BQN functionality requires seeing the subscribers’ individual IP addresses (for example, to limit each subscriber maximum rate). For that reason, it is important to deploy the BQN in a network position where there is no NAT between the BQN and the subscribers. There can still be NAT between the BQN and the subscribers, but in those cases ,the rate and shaping limitations, and the subscriber metrics, will apply to the NAT IP address. TCP Optimization will also work if there is NAT, but it will not benefit from per-subscriber adaptations.

It is crucial to make sure that traffic through the BQN is symmetric: if the downlink traffic for any subscriber is going through the BQN, then all the downlink traffic and the corresponding uplink traffic for that subscriber must go through the BQN too. Otherwise, the BQN will not be able to limit the rate or do shaping for that subscriber(because it would not see all the traffic), and the TCP optimization could even block downloads towards that subscriber (because it would not see that some traffic has been acknowledged). Care must therefore be taken if there are redundant links for those links going through the BQN, so that only in case of failure are redundant links bypassing the BQN used. If there is load balancing among the links, all the load-balanced links must go through the BQN.

Regarding whether it is better to place the BQN closer or further from the subscribers, we must first consider only places where there is no asymmetric traffic and, preferably, no NAT towards the subscribers. Then, from a TCP Optimization perspective, it is better that the most difficult hops (e.g., a low-quality wireless backhaul link) are between the BQN and the subscribers, because the TCP Optimization will then help us over that difficult hop. In fact, a BQN could be installed on both ends of a very challenging transmission hop (like a satellite link), but we should also look at deployment options that result in a minimum number of BQN nodes.

It is recommended that a bypass path is established between the neighboring nodes of the BQN (Access and Internet gateways in the diagram above), so if there is a failure in the active link or the BQN, the traffic is automatically steered through the bypass.  See the section Setting up a bypass.

‍