3 ms·

We need to get the physicists (general relativists?) working on upping the speed of light. Seriously though, how much larger is your latency than speed of light

We need to get the physicists (general relativists?) working on upping the speed of light. Seriously though, how much larger is your latency than speed of light limitations? 2x?

xoa 9y ago
TrueNet keeps track of a lot of good ISP performance measures, including one for international latency [1]. Looking at that average latency to the USA at least it seems pretty damn decent, only about 30-60% worse then the theoretical limit imposed by current fiber tech (speed of light in standard silicon glass fiber is about 30% slower then in vacuum), since direct line is about 7000-8000 miles. And of course as a practical matter fibers are not run directly between every destination, so add on extra for that and if TrueNet's numbers are reasonably accurate then everyone seems to be doing a reasonable job given the distance. I remember reading years ago about interesting research being done to use photonic-bandgap rims to make fibers that would allow propagation approaching vacuum speed [2], and if deployed eventually that could drop latency by another 40-60ms. But at least back then there seemed to be enough loss (3.5 dB/km) that it'd be impractical over long distances without excessive boosters, even ignoring any cost/durability issues of the fiber itself. That may have been improved in the last 4 years though, but I'm not sure if it ever did make it to market (or if there are even plans still in progress for it, raw throughput was impressive too FWIW). At any rate: the NZ ISPs seem to be offering pretty solid service on that front even internationally, at least in terms of pure latency. If there is increasing congestion or other issues those are just matters of capex, and between the ISP's servers and premise links there might be more lag being added in. But I don't see any glaring core inefficiencies there in terms of the international link part, unless I'm missing something entirely. Looks a lot better then I'd have expected taking a guess based on how it was a decade ago. ---- 1: https://truenet.nz/international-latency https://truenet.nz/international-latency 2: "Towards high-capacity fibre-optic communications at the speed of light in vacuum": https://www.nature.com/articles/nphoton.2013.45 https://www.nature.com/articles/nphoton.2013.45 (DOI: 10.1038/nphoton.2013.45)
- ioquatix 9y ago
  So, I do understand there are limitations, but I guess I think that once the data has started to flow, it should maximise available capacity. However, I feel like `pacman -Syu` should be faster. I don't mind the latency, but the throughput should be better right? I only get 1-2 Mbyte/s average using US servers.

We need to get the physicists (general relativists?) working on upping the speed of light. Seriously though, how much larger is your latency than speed of light limitations? 2x?

xoa 9y ago

TrueNet keeps track of a lot of good ISP performance measures, including one for international latency [1]. Looking at that average latency to the USA at least it seems pretty damn decent, only about 30-60% worse then the theoretical limit imposed by current fiber tech (speed of light in standard silicon glass fiber is about 30% slower then in vacuum), since direct line is about 7000-8000 miles. And of course as a practical matter fibers are not run directly between every destination, so add on extra for that and if TrueNet's numbers are reasonably accurate then everyone seems to be doing a reasonable job given the distance. I remember reading years ago about interesting research being done to use photonic-bandgap rims to make fibers that would allow propagation approaching vacuum speed [2], and if deployed eventually that could drop latency by another 40-60ms. But at least back then there seemed to be enough loss (3.5 dB/km) that it'd be impractical over long distances without excessive boosters, even ignoring any cost/durability issues of the fiber itself. That may have been improved in the last 4 years though, but I'm not sure if it ever did make it to market (or if there are even plans still in progress for it, raw throughput was impressive too FWIW). At any rate: the NZ ISPs seem to be offering pretty solid service on that front even internationally, at least in terms of pure latency. If there is increasing congestion or other issues those are just matters of capex, and between the ISP's servers and premise links there might be more lag being added in. But I don't see any glaring core inefficiencies there in terms of the international link part, unless I'm missing something entirely. Looks a lot better then I'd have expected taking a guess based on how it was a decade ago. ---- 1: https://truenet.nz/international-latency https://truenet.nz/international-latency 2: "Towards high-capacity fibre-optic communications at the speed of light in vacuum": https://www.nature.com/articles/nphoton.2013.45 https://www.nature.com/articles/nphoton.2013.45 (DOI: 10.1038/nphoton.2013.45)

ioquatix 9y ago
So, I do understand there are limitations, but I guess I think that once the data has started to flow, it should maximise available capacity. However, I feel like `pacman -Syu` should be faster. I don't mind the latency, but the throughput should be better right? I only get 1-2 Mbyte/s average using US servers.