Application performance was poor and the infrastructure was blamed. A six-figure switching refresh was being discussed because the symptoms appeared to sit somewhere in the network. That is a common pattern: when a system feels slow, hardware becomes the easiest suspect.
The existing access switches were not new, but they were not at end of useful life either. They still had years of support and the observed load did not justify treating them as the primary cause. Replacing them would have been expensive, disruptive and convenient for everyone except the client.
Traffic analysis and packet captures showed a different story. The problem was a long-standing configuration bottleneck in the path between systems. The design had tolerated it for years because the estate had changed gradually around it. The bottleneck only became obvious when demand increased.
The remediation was small: correct the path, remove the constraint and validate the result. The hardware stayed. The performance issue disappeared.
This is why infrastructure decisions need evidence before they need budget. A refresh may still be the right decision in some environments, but it should be made because the evidence supports it, not because nobody has looked closely enough to prove otherwise.
Engineering lessons
- Hardware age is not the same as root cause.
- Packet captures are often cheaper than procurement mistakes.
- A small configuration change can outperform a large replacement when the diagnosis is right.