An architect must address sustained congestion on the access and distribution uplink of network. QoS has already been implemented and optimized, but it is no longer effective in ensuring optimal…

Reconfiguring QoS based on the IntServ model is generally not scalable for enterprise networks, and the problem statement implies a bandwidth bottleneck rather than a QoS model deficiency, as QoS is already optimized.

Utilizing random early detection (RED) is a QoS queue management mechanism to prevent tail drops, but it does not increase bandwidth and is likely part of the already optimized QoS configuration.

Implementing higher-speed uplink interfaces directly increases the total available bandwidth, which is the most effective solution when sustained congestion persists despite QoS optimization.

Bundling additional uplinks into logical EtherChannels (or Link Aggregation Groups) aggregates the bandwidth of multiple physical links, providing increased capacity and load distribution to alleviate sustained congestion.

Configuring selective packet discard drops lower-priority traffic during congestion, but it does not increase overall network performance for all traffic and implies that QoS is already in use for prioritization.