In reply to a post by E300:

The best test is to ping your gateway, as everything goes via that gateway, this is the minimum latency you can achieve.

One thing to note is that routers tend to prioritise traffic that flows through it, rather than to it, so the ping to a gateway might not necessarily be a lower bound.

Jumping on with more anecdotal results, I have both Hyperoptic (500/500) and Openreach FTTP (40/10) at my flat in London.

Pinging Google DNS (8.8.8.8) on Hyperoptic, I get

--- 8.8.8.8 ping statistics ---
1000 packets transmitted, 1000 received, 0% packet loss, time 140ms
rtt min/avg/max/mdev = 0.520/1.451/17.573/2.265 ms

And on Openreach, I get

--- 8.8.8.8 ping statistics ---
1000 packets transmitted, 1000 received, 0% packet loss, time 624ms
rtt min/avg/max/mdev = 3.009/3.947/21.666/0.650 ms

A traceroute with Hyperoptic shows that I stay within Hyperoptic's internal network and end up in Google's network at 209.85.248.229, and on Openreach I'm routed via the Poplar exchange, Telehouse, Volta, and then LONAP.

Subjectively, I can achieve much lower latencies on Hyperoptic, but the Hyperoptic connection tends to be under contention a lot more, resulting in much more variable latency, and the occasional packet loss.

And if you want to look at BQM graphs, Hyperoptic and Openreach FTTP - ignore the spikes every 30 minutes, that's due to bufferbloat when speedtests are ran.

My experience with Hyperoptic is that if you aren't accessing something that is on a network they are peered with then you go out over absolutely garbage quality transit. IPv6 last time I checked was all HE which was terrible.

In reply to a post by jscheah:

Pinging Google DNS (8.8.8.8) on Hyperoptic, I get

What is the difference like for something that is not anycast? e.g. bbc.co.uk

I'm on a mobile 4G connection at the moment, but:
Ping statistics for 151.101.0.81:
Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
Minimum = 19ms, Maximum = 57ms, Average = 30ms

In reply to a post by gary333:

Unless people are measuring ping to the exact same server then I don’t get how anyone can be so confident they know how much their ping varies. Things like google dns could be spread across multiple data centres.

The vast majority of servers I ping that are located in London give me a roughly 14ms response.
No matter what I ping I cannot get a response below 13.7ms.
Several servers give me a 13.7ms response.
From that I can conclude my minimum latency is 13.7ms

I'm in the Edinburgh area, so that's about right.

For example I get a 11ms ping on Virgin media to TNP in Manchester. I presume they have something to do with Virgin media as every other server in Manchester (via Speedtest.net) ranges from 19-29ms), with the majority being mid 20’s. Manchester is about 50 miles from me, yet servers around the corner can’t get anywhere near 11ms that I can get for this single company’s servers in Manchester. Servers a quarter of the distance from me still take mid 20’s. London is 150 miles away yet I get a ping of 18ms via structured communication.

Virgin Media have Speedtest.net servers "on net", within their network.

I could get 7ms pings from my home to Virgins Edinburgh test server.
However it appears ALL of Virgins traffic goes through London.
Apart from the Virgin Speedtest servers, all of my Virgin traffic had a minimum latency of 20ms.

Quite a few providers have similar setups.
All their traffic goes through London, but you can get lower pings from their on net speedtest servers than you can from any general internet traffic.

Virgin Media have Speedtest.net servers "on net", within their network.

Hi just wanted to let you know that virgin media removed the speedtest.net servers from their network when they released the "HUB 4" as they now advise customers to go to samknows realspeed.
I have a "HUB 3" and CS say go to realspeed and then i go it says error, presumably because its not a "samknows realspeed" enabled router. I mean it's fine after as they use the speed test results' but when engineers come they complain and say the tests are not on virgins network.

I will say however I found virgin's speedtest servers to have higher latency than some that were further away from their network.

In reply to a post by RR_The_IT_Guy:

I will say however I found virgin's speedtest servers to have higher latency than some that were further away from their network.

Virgin's domestic network is, I understand, rather complicated and segmented.

I never understood having speedtest servers on the ISP network, unless all the content people want to get to is also on the ISP network. (with CDN's some might be, but not all). For that reason I don't see the point in Samknows Realspeed.

Better to download some files from a well connected site, e.g. the UK mirror service, or a microsoft ISO, or a Red Hat image, etc.

If punters challenge performance VM can point to SKR as an 'independent' arbiter. Paints a veneer of credibility.

In reply to a post by Pheasant:

If punters challenge performance VM can point to SKR as an 'independent' arbiter. Paints a veneer of credibility.

Shame they can’t retro fit it to the Hub 3 then! That veneer is a molecule thin!

I can see the value in the RealSpeed test for very quickly being able to rule in or out the VM DOCSIS network. If you have a gig service and RealSpeed shows 300Mbps to the cable modem then it should be an open and shut case that you have a fault that isn’t your problem.

In reply to a post by jpm:

I can see the value in the RealSpeed test for very quickly being able to rule in or out the VM DOCSIS network. If you have a gig service and RealSpeed shows 300Mbps to the cable modem then it should be an open and shut case that you have a fault that isn’t your problem

Admittedly I quite like the idea of samknows realspeed for this as it helps save time trying three different pc's on different os's every time there is a problem so CS believe me.
Its like they go "there is no problem" at the start and by the end they are like "there is very huge problem sir", "I need to speak to manager to see what i can do" then they are like "I can give you exclusive discount just for you sir so you don't raise a ticket with the ofcom"
those are actual quotes from the recorded conversations with Virgin Media from a GDPR data request.

Edited by RR_The_IT_Guy (Tue 05-Oct-21 22:55:48)