0ms fibre connections exist outside London?

I have been seeing speedtests from Hyperoptic and Community Fibre on some youtube videos where they get 0ms on speedtest.net.

Do 0ms connections exist in other cities? I live in Manchester and curious if anyone has even gotten a latency like that on a speedtest. since I personally have never seen one below 8ms.

Examples, sure:

1ms speedtest.net ping to Hyperoptic

0ms speedtest.net ping to Community Fibre

Really low ping times ( < 1-2ms) are generally only possible in the following scenarios

1. On-net test. Testing to your own ISP-sponsored speedtest server located in the same facility or very close to the broadband network gateway (BNG) where your broadband connection terminates, with minimal geographic distance and network hops. This could all well be out of London, but crucially the speedtest is 'on-net' with optimal distance, routing and no need to traverse the wider internet.

2. Off-net test. Testing to London-based speedtest server if you're geographically within say 10 or so miles of central London. Your ISP would need to have peering/transit into one of the major internet exchanges like LINX or LONAP in the same facility, again with minimal hops. The speedtest server is not on your ISP network, but in the same facility above noted with optimal distance, routing and connectivity. See examples above.

3. A variation of (2) but based say in Manchester, with local internet exchange peering/transit and the corresponding speedtest server in the same or very close to the same facility. The routing would need to be in your favour so that there are no surprise routes to London!

TL;DR - In all scenarios, simplistically the closer you are to the speedtest server the lower your latency.

Edited by Pheasant (Wed 22-Jun-22 02:19:45)

Ping which happends at same data centre is not the be all and end all.
If you are gamer and making your decision upon this, dont! This dosn't show full story. It only show story at that data centre.

0ms pings? That would take some doing surly even if it is in their data centre?

You are right, it would make little difference to gamers, as a lot of servers are not even in this country.
To be honest, it still amazes me the speed this all happens, i have done a test just for a laugh to a server in Australia - Melbourne, 252ms for my signal to get there and back, that is crazy.

I remember phoning someone in Australia a few years ago and the delay was awful, and now we can do it via skype and no delay.

This is what I mean, we could do so much good in this world if we stopped fighting each other.
sorry, just had a blip.

Is a speedtest.net "ping" really the same thing as a command line "ping" command (ICMP ECHO) or a "ping" in a game? Are we talking about the same thing?

MrSaffron has comment in the past that the Thinkbroadband speed tester's ping is different to ICMP ECHO.

Partly it's a Laws-Of-Physics thing.

Speed of light in vacuum is about 300,000km/s. In fibre it's about 200,000km/s. A ping is a round-trip (there and back). Hence you'll get 1 second round-trip time over 100,000km of fibre, which means 1ms for every 100km of fibre.

That's *just* the propagation delay. Other delays are introduced by transmission, queuing and processing.

But if Manchester to London is 300km, then 3ms is the *absolute* best case you could get - and that's if you had your own point-to-point dedicated fibre all the way, which of course is not how things work in reality.

Even on a LAN you may be pushed to get much under 1ms; usually the time it takes for the target to process the ping and respond dominates.

Text
1 23 45 67 8	# ping 10.12.255.12 PING 10.12.255.12 (10.12.255.12) 56(84) bytes of data.64 bytes from 10.12.255.12: icmp_seq=1 ttl=64 time=0.711 ms 64 bytes from 10.12.255.12: icmp_seq=2 ttl=64 time=0.424 ms64 bytes from 10.12.255.12: icmp_seq=3 ttl=64 time=0.349 ms 64 bytes from 10.12.255.12: icmp_seq=4 ttl=64 time=0.433 ms64 bytes from 10.12.255.12: icmp_seq=5 ttl=64 time=0.459 ms ^C

If it were as simple as a flash of light in a vacuum and a mirror, it would take 100ms. [30,000km, speed of light 1ms per 300km]

So the routing, etc, electronics both ways and the processing at the far end took no more than 150ms.

The blink of an eye...

In reply to a post by jchamier:

Is a speedtest.net "ping" really the same thing as a command line "ping" command (ICMP ECHO) or a "ping" in a game? Are we talking about the same thing?

MrSaffron has comment in the past that the Thinkbroadband speed tester's ping is different to ICMP ECHO.

I cheated slightly in the examples. Those speedtest results are desktop based from an EAD-based DIA service which terminates at Canonbury exchange and is probably less than 5 miles as the crow flies to Telehouse North where this provider and CF both peer. So the results really can’t be much better.

I’m not sure how Ookla perform or calculate their on test ping - but it’s certainly not a loaded ping - which arguably is more relevant to the real world but generally the figures are unflattering.

In reply to a post by Thinker27:

If it were as simple as a flash of light in a vacuum and a mirror, it would take 100ms. [30,000km, speed of light 1ms per 300km]

So the routing, etc, electronics both ways and the processing at the far end took no more than 150ms.

The routing and electronics do play a part, but only a small one.

A bigger factor is that the speed of light in fibre is only about 2/3c.

London to Sydney is about 17,000km great-circle distance, so that makes 34,000km roundtrip. However, there is no direct fibre which follows the great-circle path. The traffic will be going from POP to POP in different countries, following the routes that the intervening transit providers have available, probably via USA or Asia, making a longer total distance.

250ms RTT to Australia is pretty decent.