Back to blog

ACS for ISPs: what to evaluate before you choose

By IXC ACS Team

Published in November 19, 2026

8 min to read

A guide to choosing an ACS for ISPs: the hidden cost of running your own ACS and a checklist to review before you sign a contract.

Many regional ISPs end up with their first ACS out of immediate need. The subscriber base passes a few thousand, support starts sending technicians out to change Wi-Fi passwords, and someone on the team sets up a TR-069 server over a weekend. A few months later the fleet is multi-vendor, half the devices show up in the list without accepting any configuration, and the question of which ACS for ISPs of that size actually works is back on the table.

The NOC usually makes the call, but the impact lands on the owner's books. The ACS decides whether tier 1 support closes the ticket over the phone or the case turns into a truck roll. There's also a cost that never appears in a vendor proposal: the hours your team spends keeping its own tool running.

What is an ACS and why does an ISP need one?

An ACS (Auto Configuration Server) is the server that a subscriber's router or ONU connects to over the TR-069 protocol so it can be configured, diagnosed and updated remotely. TR-069, also called CWMP, is specified by the Broadband Forum, and the current version is Amendment 6 Corrigendum 1, from June 2020.

The CPE starts the conversation. It sends an Inform to the URL stored on it at boot, after a reset and at every periodic interval, and the server uses that session to push whatever was queued. When the agent doesn't want to wait for the next Inform, the ACS sends a connection request, and that requires the server to have a route to the network where the CPE sits, including the CGNAT range.

In September 2026, Brazil had 58.2 million fixed broadband connections, 81.17% of them over fiber, spread across 10,054 providers, according to data from Anatel, Brazil's telecom regulator, up 6.2% over twelve months. Almost every one of those connections ends at a device in the subscriber's home, often loaned by the ISP. Without an ACS, the only way to touch it is through its web interface, with the customer on the phone, or by driving out there.

What an ACS for ISPs needs to deliver

An ACS for ISPs needs to keep the fleet pointed at it, read and write the parameters of every model you buy, run diagnostics from the CPE itself, make bulk changes with control and reconfigure devices after a reset.

Pointing comes first, because an ACS only manages what points to it. The URL stored on the CPE should be a domain name. If the server changes address, a DNS record takes care of it, while a fleet pointed by IP has to be repointed one device at a time. Networks with OLTs and ONUs from different vendors have their own quirks, which we cover in how to set up the TR-069 ACS URL in multi-vendor networks.

In practice, remote diagnostics means ping, traceroute and speed tests triggered from the CPE itself (ONU or router), not from a NOC workstation. If the speed test on the device matches the plan and the subscriber's phone doesn't, the problem is almost always inside the house.

Reconfiguration after a reset is one of the biggest drivers of field visits. The subscriber presses the button because the internet went down, the ONU comes back with factory settings, no PPPoE and no Wi-Fi network, and without an ACS that becomes a truck roll. Our post on automatic provisioning of ONUs and routers shows how that flow works.

Why does the router show up in the ACS but won't change the Wi-Fi password?

Almost always because the password lives in a vendor-specific branch that the ACS hasn't mapped yet for that model and firmware. The Inform carries the basics of any CPE (manufacturer, model, serial number, uptime), which is why the device appears in the list. For everything else, the ACS needs to know where each vendor put each setting.

TR-069 standardizes how the conversation happens, but it doesn't force vendors to use the standard path. Older models tend to use the TR-098 tree, which starts at InternetGatewayDevice., newer ones use TR-181, which starts at Device., and much of what support cares about, like the WAN VLAN and optical power, sits in X_ branches, such as X_HW_ on Huawei and X_ZTE-COM_ on ZTE.

Certifying a model means building that map and linking each ACS function to the right path. The map, though, is only valid for the firmware and hardware version that was tested. All it takes is the vendor moving a parameter in an update for password changes to stop working. So ask the ACS vendor which firmware each model was tested on, and before buying a batch of CPEs, buy one unit, point it at the ACS and check.

How much does it cost to run an in-house ACS?

The cost of an open source ACS built in-house is in the hours of skilled staff it eats up every month, and those hours grow with the fleet and with the variety of models.

The server depends on a database, a runtime and libraries, each with its own update cycle, and an update to the operating system or to the ACS itself can break the integration between them. Teams often react by simply not updating anymore, which keeps the server up that week and leaves packages with known security flaws running for months.

Mapping models usually takes more hours than maintaining the server, and almost nobody tracks it. Every new model requires someone to open the parameter tree, find the right branches, write the provisioning scripts and test reset, Wi-Fi and optical readings, and every new firmware can mean redoing part of that work. With three models it's doable. With twelve, someone ends up doing nothing else.

In a homegrown ACS, the scripts and per-model exceptions tend to live in one person's head, and when that person leaves the ISP nobody wants to touch the server anymore. Our view is that an in-house ACS makes sense for a fleet of one or two models with someone dedicated to it. In a multi-vendor fleet, the numbers rarely work out.

What changes with a managed ACS

With a managed ACS, mapping models, updating the server and protecting bulk operations become the vendor's job. IXC ACS, used by more than 700 ISPs, has more than 230 certified models, each with a record of the versions tested, function by function, and the known caveats. Its certification process is clear about the limit: certified does not mean 100% functional, and on firmware other than the one tested, support is expected rather than confirmed.

The bulk firmware update recommends starting with three devices and stops on its own when failures reach 30%, in batches of ten or more. Automatic reconfiguration reapplies the settings after a reset.

At one Brazilian ISP analyzed, with tens of thousands of support interactions over six months (2025 to 2026), subscribers with a CPE in IXC ACS had 62% fewer truck rolls and a 7-day ticket reopen rate 6.2 percentage points lower than those without one. That's one provider's result, and yours depends on your equipment mix and the automation you turn on.

Checklist to evaluate an ACS before you sign or migrate

  1. Which models and firmware versions in my fleet are certified, and on which version were they tested?
  2. Where can I see, per device, what works and what doesn't?
  3. How do I request certification for a new model?
  4. Is the ACS URL a domain name? Does switching servers require repointing the fleet?
  5. Does an ONU that was reset come back configured on its own?
  6. Does bulk firmware update have a test batch, scheduling and an automatic stop on errors?
  7. Who updates the server, and what happens if the update fails?
  8. Does the server have a route to the CGNAT range of the CPEs?
  9. Can the agent see the device without leaving the ERP and the customer service tool?

Frequently asked questions

What port does TR-069 use? Port 7547 is the one registered for CWMP and the one most CPEs use to receive the connection request. The port where the ACS receives the Inform depends on the installation, so confirm the one on your server.

If the ACS goes down, does the subscriber lose internet? No. The ACS is a management channel and does not sit in the path of subscriber traffic. While it's down, diagnostics and configuration changes stop, and devices reappear as they send their next Inform.

Can you switch ACS vendors without a truck roll? It depends on how the fleet is pointed. With the URL on a domain you control, migration is much simpler. With an IP, the URL has to be rewritten on every device, usually through a bulk change pushed from the old ACS while it still responds.

Does the ACS manage the OLT? No. TR-069 manages the CPE (ONU, ONT and router). The OLT stays on its own management system, and what the ACS shows about the fiber is the optical power the ONU reports.

Where to start the evaluation

Before comparing vendors, list which models and firmware versions are in the field and how many of each, because every answer you get needs to be checked against that list. Then take the model that generates the most tickets and test Wi-Fi changes, optical readings, reset and speed test on it.

If you'd like to run that test with IXC ACS on your own fleet, click the button below to talk to our team.

Let's elevate your client experience togheter?

With IXC ACS, you manage devices, run diagnostics, and perform mass operations remotely. Offer an agile and efficient support experience for your customers while saving time and resources.

    ACS for ISPs: what to evaluate before you choose