Network engineering for ISPs
What does an ISP network need to grow without outages? A stable BGP edge, a routed core, addresses to grow into and well managed queues. Our network engineering for ISPs tackles each of those in stages, without cutting service to your subscribers.
- BGP and multihoming
- IPv6 and CGNAT
- QoS
- 10+ years
LACNIC ran out of free IPv4 addresses for Latin America and the Caribbean.
Source: LACNICof users reached Google over native IPv6 on March 28, 2026.
Source: Internet Society Pulseis what an ISP running CGNAT must log to identify each subscriber.
Source: RFC 6888Problems we solve in ISP networks
These are the symptoms we see most often when auditing an internet provider's network. Each one has a technical cause and a concrete fix. They are the starting point of any network engineering for ISPs project.
Symptom
Out of IPv4 and growth is stalling
Every new subscriber means buying or sharing addresses, and CGNAT starts running out of ports.
How we fix itWe size CGNAT properly and roll out dual stack IPv6.
IPv6 implementationSymptom
Slow at peak hours even when the link isn't full
Latency climbs under load, and subscribers feel it on calls, games and video meetings.
How we fix itWe manage queues and per-subscriber plans with quality of service.
Quality of service (QoS)Symptom
One transit provider carries the whole network
If the upstream fails, every subscriber goes down and there is no alternate path.
How we fix itWe design multihoming with BGP, filtering and RPKI validation.
BGP configurationSymptom
A bridged network growing out of control
A problem at one site spreads everywhere: broadcast storms, loops and cascading outages.
How we fix itWe migrate to a routed OSPF network, one site at a time.
Bridge to OSPF migrationSymptom
Nobody knows how the network is built
There are no current diagrams or backups, and everything lives in one person's head.
How we fix itWe audit, document and set up automatic backups.
Audit and documentationSymptom
Slow provisioning and fragile authentication
Activations are manual, and a RADIUS failure knocks subscribers offline.
How we fix itWe deploy redundant RADIUS and DNS.
DNS and RADIUS serversWhat does a well designed ISP network look like?
Splitting functions into layers lets you grow in blocks, and each layer can fail without dragging the others down.
- Two internet exits with BGP keep a provider failure from becoming a total outage.
- A routed core isolates problems at each site and finds alternate paths on its own.
- Subscriber services sit apart from the core, so they scale without a redesign.
Network engineering services for ISPs
Each service has its own page with the process, benefits and frequently asked questions.
Signs your ISP network needs an audit
- Slowness complaints rise at night even though total traffic barely changes.
- A single provider or node failure takes many subscribers offline.
- The IPv4 pool is almost full and there is no IPv6 plan.
- Configuration changes go in without a window or rollback plan.
- There is no up to date network diagram.
- Network knowledge depends on one or two people.
How do we work with you?
Diagnosis
We listen to the problem and review the network through secure access.
Plan
You get findings, priorities and a clear proposal.
Execution
We apply changes in stages, with a window and a rollback plan.
Follow through
We document, train your team and stay available.
Network engineering for ISPs: frequently asked questions
What does a network engineering firm do for an ISP?
It audits the network, designs the architecture and implements it in stages: a BGP edge, a routed core, IPv6, CGNAT, quality of service and documentation. Your team gets everything explained at the end.
How much does a network audit for an ISP cost?
It depends on the number of sites, devices and services. After a free first conversation we define the scope and send you a fixed proposal.
Does my ISP need BGP and its own ASN?
If you want more than one transit provider, to announce your own addresses or to peer at an IXP, yes. We help you judge the timing and configure it safely.
CGNAT or IPv6: which comes first?
Usually both, in parallel. CGNAT buys time to keep growing on IPv4, while IPv6 takes load off CGNAT as adoption grows.
Why do subscribers complain about slowness when the link isn't saturated?
Often the cause is latency under load from poorly managed queues. Quality of service and intelligent queue management at the congestion points fix it.
How do you scale an ISP network without outages?
In stages and in maintenance windows, with a rollback plan for every change and testing with real traffic before moving to the next site.
Do you work with MikroTik, Juniper and Cisco in the same network?
Yes. We are multivendor and rely on standard protocols such as BGP, OSPF, IS-IS and MPLS so different vendors work together.
Can you work alongside my technical team?
Yes, and we recommend it. Your team knows the day to day; we bring experience from many networks and leave the know-how in house.
Do you work with ISPs in Latin America?
Yes. We work with operators in Latin America and the United States, in Spanish and English, mostly through secure remote access.
What do I get at the end of a project?
Diagrams, documented configurations, operating procedures and a report covering what was done and what is still pending.
Guides to go deeper
Other industries we serve
Is your ISP network ready to grow?
Book a conversation with an engineer. We review your case and tell you where to start.