Is fibre faster than ADSL1 for international?

Forums › New Zealand Broadband › Is fibre faster than ADSL1 for international?

adx

21 posts

Geek

#214484 14-May-2017 12:27

Sorry for the clickbaitish title, but I can't 'for the life of me' find an answer to this.

The issue is that on ADSL1 in Christchurch (~6Mbps long line to exchange) typical download speed is 600kB/s for national or cached international eg Windows updates or something from a CDN. No problem there. We can watch 2 HD streams on YT (just, but it works reliably these days).

But no matter what I download from a non-huge international site (perhaps an update for some semi-obscure software, like the Arduino tools for ESP32, or a vendor's video on some niche engineering topic), it very commonly limits at say 300kB/s or often less, pretty much any time of the day but worse during NZ peak. Again no great problem, that's fast enough for most things. Spark, if that matters.

My question is - what happens on fibre? Surely those foreign downloads (and browsing) won't still go slower than a 2001 modem's sync speed? And if they don't, what is limiting it to half that speed now?

I've asked friends and family, no answer. I've looked at Truenet, and it seems to show international speed for fibre that would be expected of a "marketing study", ie they like to show peak speed even when measuring peak usage times (ie, the fastest result of the hour), so I do wonder about their independence. Coincidentally it was around 50% of local.

I would have given my right arm a decade ago for something faster than my old ADSL1 connection, I was even bugging Telecom in 2011 about whether I could get an early connection to VDSL when it was installed in the cabinet (different place) not long after the earthquakes. Since moving in here, I just kind of gave up, for the same reason I gave up on ADSL2 after the cheap modem actually lowered the upload speed, so it was back to 2001 for me. But they've put fibre down the road now (finally...). Being a rental we'd also have "problems" with moving from copper which I won't go into, so this isn't about yes or no to fibre, but what's going on with the speeds.

kingdragonfly

11196 posts

Uber Geek

Subscriber

#1781896 14-May-2017 12:37

From "Stop the Gap"

http://stopthecap.com/2016/01/21/underseas-fiber-capacity-expanding-even-without-laying-more-submarine-cables/

"Overall submarine cable capacity, which supports a substantial amount of international Internet traffic, has grown around 36% per year for 2007-2014 and is expected to grow around 29% for 2014-2016. But traffic planners are confident the traffic growth will be easily accommodated over existing submarine cable circuits.

A new U.S. International Circuit Capacity Report from the International Bureau of the Federal Communications Commission details the total amount of capacity available between the U.S. and any foreign point. That data helps traffic planners maintain suitable Internet traffic capacity before international data traffic jams emerge. The report shows plenty of capacity remains available to handle sustained Internet traffic growth between North America and other countries around the world. Only the Pacific region, encompassing Australia and New Zealand, shows the potential for a future capacity crunch if more cable capacity isn’t introduced in the coming years.

Submarine cables laid more than a decade ago are showing vast capacity improvements, not because new fiber is being laid underwater, but because of developments in submarine cable technology.

“The technology standard has evolved from 280Mbps per pair (TAT-8 cable) in the mid-1980s, to 5Gbps (TPC-5) in the mid-1990s, to 10Gbps in 1998,” says the report. “Since 1998, the 10Gbps fiber pair has been the standard for all new cables. There are plans to deploy 40Gbps or even 100Gbps fiber pairs. Moreover, the use of Wavelength Division Multiplexing (WDM) technology can multiply the capacity from one pair to multiple pairs depending on the wavelength (or color) of the cable.”

One exceptional example comes from the Pacific region, where Internet traffic has exploded. The Southern Cross cable, which connects Australia, New Zealand, Fiji, Hawaii, and the United States, began service in 2000 offering a total capacity of 20Gbps. Those behind the project envisioned that technological advancements would eventually allow the cable to achieve a total of 120Gbps of “fully protected capacity.” They vastly underestimated what ingenuity in data transmission would bring just 16 years later.

Cross engineers are now deploying circuits capable of 40 and 100Gbps technology, bringing Southern Cross cable’s total available capacity to more than 12Tbps (12,000Gbps). Every upgrade was conducted at the cable station with zero new fiber pairs laid in the water. Other undersea cable operators are initiating similar upgrades, providing exponentially greater capacity at a minimal cost.

The report found the most popular destination for U.S. international undersea cables was Colombia, which hosts eight. Japan and the United Kingdom are each reached by seven U.S. cables. Five cables each reach Panama, Brazil, and Venezuela, and Mexico and Australia have four each.

The most aggressive capacity upgrades are scheduled for the Atlantic region, mostly to support increasing traffic from Europe, the Middle East, and especially Africa. The Pacific region, in contrast, has just 13.3% non-activated capacity, possibly demonstrating a need for new cable capacity."

hio77

12999 posts

Uber Geek

ID Verified

Trusted

Lizard Networks

#1781898 14-May-2017 12:41

all depends on the two end points.

tcp scaling sizes, latency, routing etc

Image result for tcp latency throughput

#include <std_disclaimer>

Any comments made are personal opinion and do not reflect directly on the position my current or past employers may have.

adx

21 posts

Geek

#1781920 14-May-2017 13:39

Funnily enough I was recently reading about the latency thing. I know a bit about the internals of networking protocols, but obviously not enough.

That looks very much like the problem, from the outside, and the graphs. I too would be quick to blame the cable, but I'd like to know the actual reason and I simply wasn't looking at my own interface on this machine - even the VM network interfaces are bridged so it might pay for me to go and do the testing on a raspberry pi of all things. From a quick peek, I've got the horrible feeling that I applied some "network speedup" registry hack many years ago and that has limited my TCP window size to not very big, even though receive window scaling is enabled.

Not a big fan of emojifavicons whatever they're called but I might need to insert one that looks sheep-ish.

I'll check.

adx

21 posts

Geek

#1781936 14-May-2017 15:15

Following up: I'm sold. It's the receive window too small on my machine.

Initially it was looking like that wasn't it, but all "fast" connections to international sites were to eg ftp.citylink.co.nz or akamaihd.net with local pings (22ms or so), while all the more obscure connections (eg those at http://www.arduino.org/previous-releases to reuse that example) were 320kB/s or so. No tweaking of TCP parameters altered it, including forcing the receive window to 4kB (!). I fired up wireshark and instantly saw everything's fixed (ie broken) at 64KB window length no matter what I do. Wondering if this was somehow something to do with my router, I went to a Vista machine on the same network and lo and behold the full 800KB/s* (or maybe that's kB) same downloads (made no difference whether I hit the same one, nothing seems to be cached).

* getting faster by the day as neighbours disconnect from copper

I got me some debugging to do. And my excuse to not go to fibre has both disappeared and expanded at the same time, but I'm sure it'll be easier to twist my arm now that it'll actually work.

Many thanks for the tips, I've been yanked into the 21st century finally!