Our Enterprise Network Upgrade: 3 Avoidable Mistakes (and the Checklist We Now Use)

I'm a senior network architect handling enterprise infrastructure orders for about 9 years now. I've personally made—and documented—three significant mistakes in the last two years, totaling roughly $47,000 in wasted budget and rework. I now maintain our team's pre-deployment checklist to prevent others from repeating my errors.

This article is for IT managers and network ops leads planning an upgrade or refresh, especially with Nokia equipment like the 7750 SR routers, 1830 PSS platforms, or private wireless (4.9G/LTE) solutions. If you're sitting on a pile of quotes and proposals right now, this list is for you. It covers three main pitfalls and a six-step checklist to navigate them.

Mistake #1: The 'Compatibility' Trap

In my first year (2017), I made the classic compatibility assumption. We had a mix of legacy Cisco routers in a regional data center, and the plan was to phase them out with Nokia 7750 SR-14s. I checked the interface types—all 10GbE—said "we're good," and signed off.

They arrived. They fit the rack. The light turned green. But the MPLS LSPs kept flapping. I said "our Nokia routers support standard MPLS." The existing team heard "it'll plug into the existing MPLS-TE tunnels." Discovered this when traffic engineering didn't work.

Result: three weeks of troubleshooting, a $12,000 consulting engagement to rewrite the tunnel policies, and a reputation dent. Looking back, I should have done a full protocol/feature parity mapping. At the time, I thought "same speed, same connector" was enough. It was not.

Mistake #2: The 'Private Wireless' Oversight

Switching gears—we deployed an industrial private wireless network (Nokia Digital Automation Cloud) at a manufacturing plant in September 2022. The facility was using 30-year-old SCADA systems over RS-232 serial connections. The client wanted to modernize with 4.9G LTE.

I focused 100% of my specs on the radio layer—coverage maps, capacity planning, RU placement. I literally ignored the edge compute and legacy gateway requirements. I messed that up.

I went back and forth between a virtualized Nokia Edge Gateway and a hardware appliance for about a week. The virtual offered flexibility; the appliance offered guaranteed latency. Ultimately chose the virtual because it was cheaper. But the SCADA systems spoke Modbus RTU over serial, not IP. There was no gateway configured to bridge that. The device radios connected fine to the RUs. The backhaul to the core worked. But nothing on the factory floor could talk to anything. The mistake affected a $27,000 order where every single item had to be re-architected. We discovered the error when the first test device tried to read a temperature sensor—silence.

$27,000 for the equipment, plus a 6-week delay for the custom gateway modules. If I could redo that decision, I'd spec the edge gateway hardware day one, with detailed I/O mapping. But given what I knew then—zilch about 20-year-old serial protocols—my choice was naive, not malicious.

Mistake #3: The Licensing Assumption

This one is smaller but embarrassing. We ordered a Nokia VSrx virtual firewall for a branch office. The software license model is based on throughput tiers. I ordered the '10 Gbps' tier because their internet link was 5 Gbps. That's obvious, right?

We were using the same words but meaning different things. The throughput licensing applies to *virtual* throughput in software, not wire speed. Due to the vCPU allocation we gave it, the actual throughput was around 3 Gbps before the CPU pegged. Error caught when the first speed test maxed out at 3.2 Gbps. The wrong license on 1 unit cost $380 in upgrade fees plus an afternoon of config changes.

Bottom line: never assume what a 'throughput' license actually covers until you see the official capacity planner. Ask for it explicitly.

The Checklist We Now Use

So here's the six-step checklist I maintain. It's saved us from at least four errors in the last 12 months alone.

Step 1: Protocol/Feature Parity Mapping (Before Quote)

Document every L2/L3 protocol your current gear runs. Then map it explicitly to the Nokia product's supported features. Don't assume 'OSFP' equals 'OSPFv2 and v3 with the same timers.' We caught a mismatch on BFD timers this way that would have caused flapping again.

Step 2: Physical & Environmental Audit

Check rack depth, power (AC vs. DC, voltage ranges), and cooling capacity. Sounds boring. But in 2023, we missed that the Nokia 1830 PSS shelves required 3U instead of the 2U we had budgeted. It meant a different cabinet. We caught that before ordering, but it reset the timeline by two weeks.

Step 3: Licensing Capacity Planner

Ask for the official licensing tool or guide for software-based products (VSrx, virtualized gateways). Then triple-check the throughput your actual workload needs. The 'maximum' on the spec sheet is under ideal conditions with optimal CPU. Real-world is lower. Use the official capacity calculator, not a guess.

Step 4: Legacy Integration Points (The One Most People Forget)

This is my personal 'gotcha'—everyone focuses on the shiny new 4.9G/5G or IP routing. The hardware is cool. But what's it connecting to? If you have RS-232 serial SCADA, 4-20mA analog sensors, or old POTS lines, plan the gateway *before* the radios. In a private wireless deployment, the network is only as useful as the edge gateway connecting your old stuff to the new network. Diagram the I/O flow from sensor to core. If there's a serial or analog device, list the exact protocol and connector. That's the bottleneck.

Step 5: Pre-Deployment Integration Test (Staging)

Don't ship to site first. Configure a lab rack with one of each device: router, switch, security gateway, industrial gateway. Power them on. Build one MPLS tunnel, one VPN, one device registration on private wireless. See if they talk. We found a VLAN mismatch in 15 minutes that would have taken a week to fix on site with the factory running.

Honestly, this saved us a ton of time on the last two projects. The automated config check script now runs first thing in staging.

Step 6: Change Management & Documentation Handoff

When the order's in and the LAB is stable, write a three-page summary: what we ordered, why, any deviations from the original RFP, and the exact licensing keys. Hand it to the deployment team *before* they start racking. I used to hand over a folder full of PDFs. That was stupid. A single Google Doc with a checklist is way more effective.

Final Notes

We've caught 12 potential errors using this checklist in the past 18 months. It doesn't catch everything—last month we missed a fiber connector type on a 10km link—but it saves the big ones.

A few extra things to watch for:

• Nokia's VSrx license model changed in 2024 (check current version).
• Some enterprise switches support 'fabric' mode for stacking; not all models do.
• For private wireless, check the eNodeB/gNB compatibility with your chosen spectrum band (CBRS vs. licensed).
• Prices as of late 2024; always verify current rates with your Nokia partner.

Avoid these three traps, and your upgrade will go a lot smoother than mine did.