A Network Engineer on What Changed in Telecom Since 2020
Archi4 min read
Contents
- The Network Job Changed
- What is the biggest shift in traffic since 2020?
- How did the hardware market change?
- Is peering harder than it was?
- Did fiber and fixed wireless change your planning?
- What changed in day to day operations?
- How does security work compare with 2020?
- Where does IPv6 stand now?
- Did the job get easier?
The Network Job Changed#
We spoke with a network engineer who runs backbone operations for a mid-sized ISP in the eastern United States. He agreed to answer questions on the condition that we not name him or his employer. The following conversation has been edited for length and clarity.
What is the biggest shift in traffic since 2020?#
The obvious answer is video. Video calls, game streams, and security cameras all grew, and the old assumption of mostly downstream traffic is gone. Five years ago we shaped uploads and assumed downloads would always dominate. Now a single home can push ten times more data than a small office did in 2020. We plan capacity for both directions, and we watch upload curves during storms and evening hours. We also see more latency-sensitive traffic from gaming and cloud desktops, so jitter monitoring is part of normal capacity planning.
How did the hardware market change?#
Lead times changed us more than speeds did. In 2020 we ordered a chassis from a major vendor and waited 60 to 90 days. During the chip shortage, delays went past a year for some SKUs. We started buying generic white-box switches in batches and keeping them in a warehouse. The configs are stored in templates, so a replacement box can go live in an hour. Vendor software still matters, but we no longer wait on a vendor schedule for basic builds. The white-box route means we own more of the software lifecycle, so we test more before rolling out.
Is peering harder than it was?#
It is more complicated. Transit prices keep dropping, and we buy less of it. But settlement-free peering with large content networks is harder to get. Five years ago we could request a session at an internet exchange and get it accepted quickly. Now many providers require paid peering or enforce stricter traffic ratio policies. We spend more time measuring packet loss, latency, and route preference. The price list is shorter, but the negotiation is longer. Backup paths used to be simple. Now we carry more routes and need to verify that failover does not drive latency up.
Did fiber and fixed wireless change your planning?#
Fiber is the default for new builds. Five years ago we designed new subdivisions with copper or coax and offered lower tiers. Now the construction cost for fiber is close enough that we skip copper entirely. Fixed wireless also stopped being a rural-only product. We use it for temporary events and for customers in valleys where trenching is impossible. The tricky part is spectrum and backhaul, not the radio.
What changed in day to day operations?#
The work moved from typing commands to writing and reviewing automation. We can now push a config change to hundreds of switches with one template. That is good, but a bad template fails everywhere at once. We learned to stage changes in small groups and verify each step. The person who writes the Jinja template has a higher risk than the person who used to log into each device. Change review is the most careful part of the job. It also raised the bar for documentation. A config change that looked obvious in 2020 now needs a rollback plan before it starts.
How does security work compare with 2020?#
The attack mix is different. Five years ago we filtered ports and watched for broad scans. Today the noise is targeted credential stuffing and API calls. Attackers find a new public IP within minutes and start probing it with known exploit patterns. Flow telemetry and DNS logs tell us more than a firewall rule can. We also spend more time on customer premises equipment, because home routers and small office gateways are the easiest way in. We block some traffic at the edge, but the real work is in endpoint detection and in monitoring the APIs that tie our systems together.
Where does IPv6 stand now?#
It finally works in practice. Five years ago we ran dual stack for a small slice of customers and treated it as a checkbox. Now the majority of our access traffic runs over IPv6, and the remaining IPv4 traffic sits behind carrier-grade NAT. The problems come from applications with hardcoded IPv4 addresses and from logging systems that only capture one family. The protocol itself caused fewer problems than the NAT issues we did not expect. Metering and billing systems also had to catch up, because session counts no longer match address counts.
Did the job get easier?#
No, it got different. The network is faster and more automated, but the failure modes are broader. A software update can change behavior in ways a hardware fault never would. The five-year comparison is not about speed. It is about learning to manage systems where the slowest part is often the test process, not the packet forwarding. Troubleshooting now starts with checking the automation logs before the interface counters.
More articles