The CityFibre Latency Lottery: Is Your "Alt-Net" ISP Costing You Performance?

CityFibre is transforming the UK's internet landscape, rolling out gigabit full-fibre connections that promise a new era of speed and stability. When they arrive in a new area, customers eagerly sign up, trading in unstable Virgin Media lines or aging BT connections for speeds 10, 20, or even 30 times faster.

When CityFibre is in your area you'll get a letter through the door showing all the providers available to you, wow! Every provider on CityFibre's network offers the same product (Bar some exclusive deals like Yayzi had with 2.5G and Sky currently has with 5G). But what if they don't? And by don't I mean, they genuinely actually don't.

For some light reading, in 2023 Ofcom did a look into the state of FIbre marketing in the UK, and they tended to focus more on the legacy fibre products being annoying for trying to sell FTTP services.

As someone that has been afflicted by CityFibres lackluster resilience in Scotland, I monitor my connection closely. I recently stumbled upon a glaring discrepancy: two neighbours, on the same street, with the same CityFibre line, but with a 500% difference in network latency. And yes, I am using percentages for dramatic effect.

One has a 3ms ping. The other has a 17ms ping. Shocking I know, it's about time to switch to Virgin Media right? Not quite.

This isn't a faulty line. It's a fundamental difference in network design that weirdly, if you look everyone is talking about at least on the tech subreddits and such. But in some senses you only seem to find that information I think when it's too late and you're starting to ask questions. Like me. Someone who went to college for networking, and has a Diploma in this stuff, and once was a sysadmin for a local SaaS provider. Somehow I missed this.

The Mystery: A Tale of Two Traceroutes

The evidence lies in a simple network diagnostic tool called a "traceroute," which shows the "hops" your data takes to reach a destination.

Here are two traceroutes from Glasgow to bbc.co.uk, both on the CityFibre network.

Evidence A: TalkTalk

A user on a standard TalkTalk package gets this result:

1.  glasgow.talktalk.net  (3.2 ms)
2.  as13285.net     (8 ms)
3.  scr.as13285.net (7 ms)
4.  bbc.co.uk       (6 ms)

The Analysis: Look at hop 1. The data hits a TalkTalk router in Glasgow in just 3 milliseconds. The traffic enters the wider internet locally before heading south.

Evidence B: The "Alt-Net" ISP (Olilo)

A user on a 1.2Gbps package from Olilo (a smaller, alternative network or "alt-net") gets this:

1.  lon.core.olilo.net      (15 ms)
2.  gtt-transit-lon.net   (29 ms)
3.  ae0.rt-lon.core.bbc.net (17.8 ms)
4.  bbc.co.uk               (17.5 ms)

(Note: IP addresses and router names have been simplified for clarity.)

The Analysis: The very first hop outside the home network takes 17 milliseconds and is already at a router in London.

Both users are in Glasgow. Both are on CityFibre. So why is the TalkTalk user outside their network in 3ms, while the Olilo user's data has to travel 400 miles to London just to get started?

And yes, it is important to clarify, any services not accessible via the Glasgow POP would incur the same latency as it probably has to go to London anyway, but it's at least nice to know there's SOMETHING there for you to access. IE: Peer to Peer gameplay would certainly be more responsive.

The Open Secret

The answer is network architecture, and it all comes down to where your ISP builds its "on-ramps" to the internet.

Think of CityFibre as the company that builds all the new local roads and the slip-road to the motorway. The ISP (TalkTalk, Olilo, etc.) is the company that owns the motorway itself.

The "on-ramp" to this motorway is called a Point of Presence (POP).

The Big Boy Model (TalkTalk, Sky, BT): These huge ISPs have been around for decades and have the budget to build on-ramps (POPs) in most major UK cities. When you're in Glasgow, you use the Glasgow on-ramp. Your ping very low.
The "Alt-Net" Model (Many Smaller ISPs): It's expensive to build 50 on-ramps. It's far cheaper to build one or two in London (e.g., in the Telehouse North data centre).

Because their only POP is in London, all their customers from Glasgow, Hull, or Cardiff must have their traffic "backhauled" on a private link to London first.

Your data is forced to travel 400 miles south, just to be sent 400 miles north again to reach a server in your own city. This is known as the "Trombone Effect," and that round-trip is precisely where your 14ms+ of extra latency comes from. If you and your friend were both on the same network, next door to each other. You'd be around 30ms+ just to connect to each other. If me and my friend are on Sky Broadband, we're able to connect to their Scottish POP wherever it may be, and get extremely low latency. Almost LAN at that point due to how fast it is 2-3ms. That does have an impact.

Private Links

For any customers using an alt net like Olilo you probably noticed an increase in latency recently, not much, just about 3-4ms. It took me from 12-14ms to 15-20ms.
But it's noticeable, when I asked Olilo about this problem they confirmed, they hadn't changed anything and assumed it was a CF issue, CF decided promptly to ignore their complaint because it wasn't that big of a deal.

(I will say the support with Olilo has been fantastic, I just feel bad that so many problems are out of their hands).

When Olilo has a problem, it doesn't exactly have the flexing power a bSkyb has if you remember they went by that name once. A company that can and will eat the turnover of CityFibre in an afternoon.

The problem actually lies in that private "middle mile" circuit, the 400-mile data pipe from the Glasgow exchange to the ISP's London data centre. This circuit is often leased from a third-party wholesale provider.

If that primary, most direct fibre path gets cut (perhaps by a construction crew on a sunny day, or an Englishman with a superb lawnmower), the provider's network will automatically reroute all that traffic onto a backup path. This backup path, it's a backup. There's no telling what it is, what standard it runs to and who even put it there or whether it's offloaded to another third party!!

Suddenly, you are the victim of a very real fault, but one that neither your ISP (at their core) nor CityFibre (on your street) will claim. Because your service is still "working," it's not considered a high-priority outage. You are simply stuck with a permanently slower connection, a casualty of a "wholesale blame game" where no one takes responsibility for the middle mile. I'm only on the 1.2G package, but I've got pals on the 2.5G package and they claim their speeds have been cut pretty dramatically due to this change. Do you see it yet? Internet effectively in the UK has just became a bunch of middlemen pointing at everyone else. I don't think the Internet can work in any different way to be frank, but I think our national infrastructure needs some more care and attention for the love of god.

Now you might ask. But dude writing this article. Why isn't TalkTalk affected then! Haha!

TalkTalk, again is a massive provider, they aren't using the private U-Haul that CityFibre or some other provider is leasing out to people, they will have their own config, that for good or ill, will serve their needs and may go down just as Olilo stays up. It's all relevant.

Does 14ms of Extra Ping Actually Matter?

Let's be clear: for 90% of what you do, this extra 14ms is completely unnoticeable.

Streaming (Netflix, YouTube): Makes no difference. The video buffers seconds or minutes in advance.
Web Browsing: Negligible. A page might feel a fraction of a second less "snappy," but you wouldn't notice.
File Downloads: Makes zero difference. This is a measure of bandwidth (Mbps), not latency (ms).

But for me, it's a deal-breaker.

Competitive Online Gaming: YES. This is the one application where latency is king. But it only impacts if you have servers in Glasgow. And for a lot of games that won't be the case, to be frank. You'll still be going down to London and back again sadly. There's a reason competitive events are all done locally.
Video Calls & Remote Desktop: A lower ping will make live conversations and remote work feel more "instant" and less laggy. If you asked the average user would they take a 14ms delay in whatever it was they were doing. They'd say no. And people aren't told about this upfront.

A lot of the improvements that you will see, is based entirely on what your ISP is doing at the local level. Sky Broadband has a Scottish POP, but they don't peer it with LINX Scotland, so I still have to go to England and back to get results. Whereas TalkTalk and Vodafone do. Do you see how messy this all this? And that's not to say LINX Scotland is the solution, far from it. There's just no.... advertising of any of this.

The Transparency Problem (And How to Check Your ISP)

This isn't a CityFibre problem or even an Openreach problem (the same issue exists on their network). It's an industry transparency problem.

ISPs are allowed to sell "900Mbps on CityFibre" as if it's a single, identical commodity. Look at anyone dong network comparrision or "What ISP is the best" graphs and tables, they never include any information about POPs or network configuraton. Olilo is one of the first I've ever seen to mandate explaining what peering they do. The alt-nets, which often pride themselves on being transparent with "no CGNAT" and "static IPs," are strangely silent on their core network design. Olilo advertises theirs with glee, they are one of the better ones, but you still don't actually get this information from their landing page. You find who they're peered with, but I've been talking with their support team who either didn't understand me or wasn't aware of the networking consequences of said decisions. (They followed up after with a clarification which I think was great of them). I think honestly providers should be forced to provide latency numbers for at least the 4 countries in the UK. Ofcom enforced.

You are being sold a package sometimes without a key technical specification, forcing you into a "latency lottery."

Want to see your ISP's real network path? It's easy.

On Windows: Open Command Prompt and type tracert bbc.co.uk
On Mac: Open Terminal and type traceroute bbc.co.uk

Look at the first 2-3 hops.

Is the time low (2-5ms)? Does a router name include your local city (e.g., gla, man)? You're on an ISP with a local POP.
Is the time high (15ms+)? Does the router name include London (lon, thn)? You are being backhauled.

It's important to note you should always clarify this with someone else that might know what they're talking about before you go accusing your ISP of doing you dirty. If they even are. It's pretty normal to do.

Before you switch ISPs, ask their technical support one simple question:

"Where is your nearest network Point of Presence (POP) to my postcode?"

If they can't or won't answer, you have your answer. You might be getting fantastic customer support and a great price, but you're paying for it with a 400-mile detour. Heck you might also be getting a fantastic network configuration, with local POPs, and an absolutely horrible support team that still wouldn't tell you that information.