Access Closet Uplink Congestion After VLAN Consolidation
A two-building campus runs 24 access switches, each with a single 1 Gb/s fiber uplink to a distribution pair (no LAG). The access layer carries user VLANs 10-40 plus a voice VLAN 50. Peak-hour monitoring shows the uplinks sustaining 85-95% egress utilization with visible drops on the 1 Gb/s interface counters. A proposal is on the table to move from per-floor VLANs to a single flat VLAN 10 for all users, arguing that inter-VLAN routing at the distribution is the bottleneck. The distribution pair currently routes all VLANs with a first-hop redundancy protocol; the core is a routed Layer 3 fabric. Budget allows either new optics for a second uplink per access switch or a distribution upgrade, not both this year.
1. Evaluate whether flattening to a single user VLAN addresses the observed congestion. What is the actual constraint and what should be measured to confirm it?
Read the worked answer
Flattening to one VLAN does not reduce the bytes on the uplink; it only removes inter-VLAN hops. The symptom is egress saturation on the 1 Gb/s access uplink, which is a bandwidth/oversubscription problem, not a routing problem. Inter-VLAN routing at the distribution consumes distribution CPU/ASIC capacity, not access uplink capacity. Confirm the real constraint before spending:
- Check interface egress queue drops and
show interfaceutilization on the uplink itself during peak. - Verify whether traffic is north-south (to core/servers) or east-west between access switches (hairpinning through distribution).
- Look at oversubscription ratio: 24 ports at 1 Gb/s to a single 1 Gb/s uplink is 24:1.
If uplink egress is the constraint, the fix is more uplink bandwidth (second fiber + LAG) or moving heavy flows closer to their consumers. Flattening VLANs would also enlarge the broadcast domain, worsen convergence blast radius, and complicate ACL/segmentation policy - a net negative with no bandwidth gain.
2. Given the budget constraint, recommend the sequencing of remediation and justify it against the measured data.
Read the worked answer
Sequence by measured impact per dollar:
- Add the second uplink and form a LAG on the highest-utilization access switches first (those sustaining >85%). Two 1 Gb/s links give 2 Gb/s aggregate and, with a hash that spreads flows, relieve egress pressure. This directly attacks the confirmed constraint.
- Re-baseline after the LAG change; if distribution CPU from inter-VLAN routing is also high (measure it), then schedule the distribution upgrade next cycle.
- Defer the VLAN-flattening proposal - it addresses a different bottleneck and increases risk.
Justification: spend on the confirmed bottleneck (uplink bandwidth) before speculative fixes. If post-LAG data shows the distribution is now the limiter, the upgrade is justified next year with evidence rather than assumption.