By Simon Woodhead
In recent weeks I’ve poked fun at the dinosaurs in our sector, got cross at the lies and hypocrisy amongst others as well as ridiculed the fake “get global” people. I want to return today to “get global” because there’s a darker side to it and whereas we looked at it from the perspective of UK CPs using “get global” to service far-flung shores, there is also a tendency for UK CPs to use “get global” within the UK – after all, they’re a global “carrier”, spin a very good yarn and seem to make those with more fragile egos feel very important.
There’s an inescapable fact in the universe and that is the speed of light. It doesn’t matter how fast a machine you build, you cannot go faster than the speed of light. Mess with that, you mess with time, and you don’t want to do that. When designing or consuming a network this reality comes home in terms of latency, in that even at the speed of light, distance takes time. The beauty of the Internet is that you don’t need point-to-point optical circuits between every computer and every other on earth, because traffic is routed and can find an optimal path; but that takes time too.
In the world of real-time communications, time matters. It matters that you get words in order, it also matters that you get them within a reasonable period for the brain to process. That period tends to be 160ms – more than that and it’ll be one of those conversations where you’re speaking over each other or it’ll just feel odd. The less it is though, the more comfortable your brain will be as in an in-person chat you’d be a metre or two from the person speaking.
160ms is huge – that’s like the UK to Asia. Even the UK to West Coast US is only 120ms. But remember, the goal is lower, not pushing limits. Furthermore, international connectivity is the most expensive (laying fibre on the sea bed isn’t something you can casually decide to do), and by virtue of that and the business models in play, more likely to be congested than, say, a fat pipe between two local ISPs in the same datacentre in London. It also involves more routers, each of which adds latency or may have congestion of its own – think airports with connecting flights in and out, where either flight or the airport can delay your trip.
I’ve always taken a slightly different view to convention here in that I don’t think latency matters quite as much as people make out, once you’re below the threshold of 160ms. What matters far far far more in my humble opinion is jitter, i.e. the variance of latency. Jitter causes packets to arrive out of order and be discarded, i.e. chop in your conversation. Jitter buffers combat this by, you guessed it, increasing latency. You don’t really get jitter on a point-to-point circuit but routing through lots of routers over different paths with different physical characteristics and different traffic levels on networks with different business models and thus different traffic management standards is going to cause jitter. Jitter is going to destroy your conversation beyond minimal levels. So we do get to pretty much the same place – distance is bad for quality and stability.
This is one reason why we discourage CPs from routing over the public Internet, and greatly favour local and direct connectivity. It is also why even in an island as small as the UK we have 3 Availability Zones. Yes, this is for high availability, but it is also for quality – Manchester is only 7ms from London but that can matter when an excavator driver gets particularly excitable in Bradley Stoke and half of everyone’s national ring is severed – the customer who connected to us locally in Manchester or London won’t notice, the one who is now fighting for priority amongst cat pictures and porn on 50% of the capacity that was perhaps already being sweated before, will notice.
So when you’re considering a carrier, proximity matters and even if you’re not a network engineer, you should ask all of your carriers, and “carriers”, for a network map so you can assess this for yourself. And maybe ask about failover, recovery etc. because shit happens and it will affect you when it does.
There is a reason why Amazon, Google, in fact any tech giant you care to mention, deploys infrastructure locally here in the UK. They know it matters and that’s predominantly for usage massively less sensitive or valuable than real-time communications. So what do the “get global” crowd know that they don’t? A secret way to run the most sensitive traffic from a foreign country without increasing latency, jitter and risk? I’d suggest there is no secret, they’re just charlatans and writing cheques you’re going to have to cash.
Now, we have customers who are overseas; quite a lot of them. So I’m obviously not saying everyone who is outside the UK using UK services is a charlatan, but I’d hope none of them are marketing themselves as UK domestic carriers when they are not even in the UK. One beauty of VoIP is that you can backhaul traffic internationally, mesh an overseas call centre into your onshore UX etc. etc. There are also things we can do to ensure the media flows directly.
Use case matters too – a Mexican carrier taking UK numbers for a Mexican customer on a Mexican platform feels ok to me, not least because if they need emergency services they’re most likely to be Mexican! With our infrastructure inside the UK, you can configure contingencies for routing calls within the UK in the event of a loss of international connectivity. Normally have your shipping calls answered by your Mexican warehouse? Configure them to route to the UK HQ or another site if there’s an issue. You can’t do that if your “carrier” is pretending to be a UK PECN and doesn’t get the call in the first place..
That is a very very different situation to your whole UCaaS business being in the UK and you buying from a “UK carrier” and who has zero infrastructure in the UK and all your traffic is going to have to hairpin through overseas switches. That isn’t terribly helpful when one of your customers dials 999 from within the UK, to get UK emergency services, but the call hairpins through Belgium for <reasons>. Whilst that might work most of the time, the probability of it not working is always going to be higher than were it entirely on-shore and it is the choice made against that probability you’ll be asked to explain in retrospect.
Look at it another way: why the holy frick should a nan in Glasgow be unable to call an Ambulance coming from Glasgow, via the emergency call handling agency in Glasgow just because of some kind of issue in Brussels? You’re right, that should never be allowed to happen! But when it does, and the nan is your customer of your service, reliant on this “carrier” who rolled out the red carpet to you, how will you explain your way out of a 10% of relevant turnover fine levied on you? And let’s just all hope the nan doesn’t die!
Who benefits from this? It isn’t you – you take on more risk. It isn’t the nan who needs an ambulance – she is at greater risk of not getting one and dying. It isn’t the UK market – that just got dirtier. Ah yes, the charlatans offshore profit and loss is the only beneficiary because they can charge as if they have UK infrastructure, lie about being a UK carrier to justify it, but just eliminate the cost. Wonderful business model if you’re immoral. It is similar to portraying your boring old US carrier business as a CPaaS provider to deceive the market into giving you a CPaaS valuation, just with more lives at risk.
And guess what? For once the government agrees with me! In the Telecommunications Security Regulations, Regulation (3)(3)(f), which applies to any provider of a Public Electronic Communications Network, defined in section 32 of the Comms Act so broadly as most likely include you if you’re a CP, and definitely to include anyone passing themselves off as a “carrier” states (emphasis added):
The duty in paragraph (1) includes in particular a duty:
[..]
(f) to take such measures as are appropriate and proportionate to ensure that the network provider:
(i) is able, without reliance on persons, equipment or stored data located outside the United Kingdom, to identify the risks of security compromises occurring,
(ii) is able to identify any risk that it may become necessary to operate the network without reliance on persons, equipment or stored data located outside the United Kingdom, and
(iii )if it should become necessary to do so, would be able to operate the network without reliance on persons, equipment or stored data located outside the United Kingdom.”
That seems pretty clear to me. It doesn’t say it is ok to pretend to be in the UK. It doesn’t say it is ok to lie about being in the UK. It doesn’t say take the margin as if you’re in the UK but save the cost. It says, pretty unambiguously, that if you are a UK PECN you need equipment in the UK such that you can deliver service even if UK international connectivity is jeopardised. That could be because Mr Putin has been playing with his underwater scissors, or it could be because of some enthusiastic Belgian (I know, just go with it) construction worker. The greater the distance, the greater the risk, period.
So, based on your carrier choices, in the event of a loss of all UK international connectivity, do you have the capability to be able to operate your services? Are your services even lawful today?
If you want to answer those questions, your first port of call is your “carrier” – ask them for a network map and failover/recovery plans. You might also find the tools at https://stat.ripe.net useful – pop in an IP address (perhaps the SIP proxy / SBC they gave you) and it’ll give you all sorts of useful information, including geolocation. Geolocation isn’t always accurate so good old traceroute is my go to.
When you discover they only have equipment in Brussels and Frankfurt or Dublin or even Ashburn, and nothing in the UK, you should give us a call (which won’t be tromboned thrice around the world)
