If you want to be fair to ISPs, don't look at cases where they fail to do the impossible or very hard, look at cases where they fail to do the very easy.
The standard deviation of any ping test I do is at least 3ms due to how DOCSIS works, and I can't ping my next door neighbor (same ISP) through the internet in under 16ms. I've done traceroutes against several servers that are in the same city as I am in North Carolina and the only one where my traffic didn't first go to Atlanta or DC (or both) was to a server hosted by an ISP that has no physical presence outside of North Carolina. My cable modem and my ISP's CMTS each have out of control bufferbloat that adds hundreds of milliseconds of latency under load in each direction, which can't be entirely mitigated by my router's traffic shaping and AQM. There's little reason to believe that they've got got any AQM further upstream given the large latency spikes I see even when my last-mile link is quiet. Disregard for latency is pervasive in the design of the ISP's network.
This analysis is very much wrong. So many people don't use their ISP's DNS servers, especially in Europe/Asia. Physical proximity does not guarantee improved latency, and many times latency varies to/from the same node, based on traffic.
That's exactly GP's point, right? No matter us-east-1 or not, what makes things slow has nothing to do with DC location. That latency is insignificant.
> Latency to where?
Greyface also asks "Latency is an end-to-end metric which will almost always involve path components beyond the control of your ISP - can and should they be held responsible for those paths?"
The latency you see is almost always from last-mile software, under the control of the ISP, and can be fixed locally. Before fixing, my local ISP gave me ping-time to the internet interconnect point in downtown Toronto that were typical of a link to Istanbul, Turkey (:-))
They weren't trying, and aren't to this day. Arguably they need a nice unfriendly regulator to require at least a good-faith attempt.
I'd invert that - by far the lowest latency ping I've seen for an external service was fly.io (single digit ms), and I never worked out quite where it was.
You must be very close to a large datacenter that houses the CDNs used by those large websites if you are seeing sub 5ms latency.
For any of the users we are talking about (underserved populations in areas without fiber internet), they will never see those latencies even with the same fiber you have.
If you live in rural Wyoming, there are going to be practically zero services hosted within 500 miles of you. You are not going to get 5ms latency no matter what you do.
If it's to the nearest IXP the ISP could create it's own IXP and measure latency to that. So it should be to not just any IXP, but a well-connected one
People like this shouldn't be writing technical articles.
A 700ms ping would be indicative of severe network problems, especially at that distance. The author is probably experiencing packet loss. Realistic ping times across the country are usually 100ms or so. I can ping most major websites in under 20ms.
DNS can indeed be a bottleneck, but it usually isn't. It's a fairly basic service, and if it's really a problem it's simple stupid to switch to a free provider like google.
Streaming video is not dependent on latency, only throughput.
Dark fiber won't necessarily help. The bottleneck is more realistically in the switching equipment. A single fiber can carry massive amounts of data.
Geographic locationa affects far more than just latency. Even though you may have a 100 Mbps link on both client and server , only 10 Mbps might make it across the Atlantic Ocean. Connections between countries/ISPs are not unlimited in capacity!
reply