Ever wondered why the upload speed is about 10 times lower on your cable modem than your download speed? It all has to do with a specification called DOCSIS. It’s what most cable internet providers use as a standard. The standard has come a long way since it’s introduction in the '90s but upload speeds were always way slower than download speeds. This article shows that that’s about to change. Welcome to “Full Duplex communication”. This is very good news for the SAFE Network. The network is partly limited by the upload speed of the users. Imagine that you as a farmer (on a cable modem) could upload 10 times the speed instead of the technology we have today. It will help the downloader on the other side to get it’s chunks way faster.
This looks to be for Hybrid Fibre Coax (HFC) networks, which were originally built for CableTV. In Australia we have the Foxtel Network which is HFC…so good news for those that have that service running down their street.
The vast majority in Australia are connected via ADSL (over copper) at present and so this technology is irrelevant.
The Australian National Broadband (NBN) initiative however, is utilizing HFC for sections of the network and have just completed testing:
On our HFC pilot, we were using the current generation DOCSIS 3.0 technology, but from mid-2017 we plan to deploy next-generation DOCSIS 3.1 technology which is capable of wholesale downstream speeds of 10Gbps and wholesale upload speeds of 1Gbps.
It seems that with the advent of DOCSIS 3.1, areas with cable available may be side stepped for a fibre rollout.
Myself, I live in the CBD of my states capital city and if I want PayTV, I have to install satellite…crazy. Looks like I’ll end up with wireless or fibre when the NBN finally arrives.
This is the current plan for the NBN and I believe some contracts were made with Telstra for use of their HFC. Not sure about Optus. Its mainly Melbourne, Sydney and Brisbane that will have NBN over HFC.
I heard that contract was awarded for the equipment install for the new DOCSIS 3.1 Telstra and Optus have DOCSIS 3.0 already and have been supplying 100Mbit/sec down and 2 Mbits/sec up for some time now.
Holland is quite small as you might know. So broadband internet penetration (especially cable) is quite high. I’m on DOCSIS 3.0 right now being able to download at 200Mbit/sec. and upload around 20Mbit/sec. It is indeed Hybrid Fiber Coax. I think it’s more cheap for the ISP to let the last few miles be cable instead of making it 100% fiber. Synchronous internet is a good thing for SAFE. I already notice that when I download an episode of a certain series every week () it’s around 1 GB. And when I’m done watching and check uTorrent I’ve uploaded about 5 times that amount before I close the torrent. That’s how SAFE should work IMO. Just running in the background providing stuff to the network. And when 1 user is quite fast it will support a user in India for example who has a very slow connection.
In Europe Liberty Global is a very big player serving 55 million users with cable. They probably feel the heat from fiber internet:
That said, Fries also sees plenty of opportunity for greater “organic” growth in Europe, where it plans to roll out the new DOCSIS 3.1 spec aggressively over the next three years.
This is all great news for us as users. Cable providers used to be monopolist parties (still are a bit). But now we see fiber (for example Google Fiber) bringing competition to the game.
Was your HFC rolled out originally for PayTV, internet or both? I know here, originally the signalling was analogue (for PAL TV) and changed out to digital within a few years.
I think the temptation will now be to utilize the HFC networks as much as possible. A large part of the cost for fibre, must be the PIT and Pipe infrastructure and I know from first hand experience, there is zero capacity to fit another cable down the POTS pipes (and lead ins)…so massive new pipe laying required for fibre.
Yes, it was all meant for just television. I still remember as kids we used to ask each other late '80s and begin '90s “Do you have cable??”. It was the end of using antenna’s on the roof. It seemed to come out of nowhere and was almost instantly there (although that’s how I remember it). The fact that Holland is so small and we all live very close to each other made it relatively cheap for the counties to implement the technology.
In Belgium also a lot of cable: a monopoly market.
Also a lot of ADSL, not much fiber.
I’ve heard in the Netherlands there is more fiber.
I also have a couple of weeks a DOCSIS 3 modem: 190 Mbps download, 11 upload, according to speedtest.
And In Australia we are still treated as the backwater of internet and while we have DOCSIS 3 the 2 monopoly Telcos give us 100Mbs down not 200 down and only 2Mbs up. Wished the new comer TPG had been large when HFC was being laid, we might have had real competition.
NBN is still only going to be 25 to 100Mbs down and 10 to 40Mbs up even though DOCSIS 3.1 will be able to do much greater.
The problem is that the 2 large Telcos are using their strength to keep the infrastructure down, including the under ocean links to the world, which they control. And even with higher speeds we would still only achieve slower speeds, since almost all our traffic is to overseas servers.
The 200 is a more expensive subscription. You have also the choice of e.g. 100Mbps, which is cheaper in Belgium.
Also, independently of monopoly Telcos, Australia is geographicly more seperated, I think → more latency in the future global Safe network, it seems to me: https://www.dotcom-tools.com/internet-backbone-latency.aspx.
But not as bad as we have now where 99.9% of our traffic has the initial latency hop from Sydney to LA USA then where ever.
SAFE will see a portion to other areas in AU, and to ASIA (nearly 3/4 worlds population) and to USA/Europe (nearly 1/4 worlds population)
So latency will be better that expected and the USA people will actually have worse than expected because while they feel they are the centre of the universe they will find out that Europe, ASIA, Africa have many more time the computers than they do.
Mind you actual latency will be fickle because of the hopping around the world each chunk will take.
Interesting point regarding density of computers, but I still don’t quite understand the relationship of XOR closeness and physical/ latency closeness.
I really would have thought placing vaults in locations with pervasive peering links across the continent would have an advantage in response times and possibly transit cost…regardless of XOR closeness.
XOR gives us randomness in normal IP space. So we can expect a random IP address for 2 close nodes in XOR space. Thus when a chunk travels from vault to your PC the route will be random in IP space.
IP space though has its densities and the higher the density then higher the probability that 2 close nodes will fall within that space. This also translates into more often the latency for routes being lower if your IP address density in your location is higher.
For instance lets say that there is only two physical locations for nodes a long distance apart. “A” & “B”. Now location “A” has 1,000,000 PCs with their corresponding IP address and location “B” has only 1,000 PC’s with their corresponding IP address.
Now for random IP addresses in routes then randomness says that any one hop will have a node occurring 1 in 1,000 hops located at location “B”. So for routes with 20 hops probability say only 1 in 50 routes will hop into location “B” when requester is in location “A”. If the requester is in location “B” then there is always two or more hops across the void (the request and delivery). Big difference in chunk delivery latency.
We can see from this extreme example that if there is one region in the world which has a significantly higher density of nodes then their latency will be lower than for a region with lower density.
Australia is only far from north America, it is a lot lot closer to the ASIA region and we actually have links that can split off in a few Asian countries. Just at the moment 99.9% of our traffic is destined for the USA with the highest density of Servers/hosting companies.
ASIA has majority of people with PCs in the world followed by Europe, then North America, followed by the rest of the world.
If people in Asia adopt SAFE then its very likely that Australia being part of Asia (links) will have good chunk latency times compared with Europe and America will have the worse out of the 3 major regions.
Although initially it seems that the highest density of nodes will occur in the USA and west europe
Disregarding the cost of hosting, I would have thought having a farm located with an ISP who had direct peering links to all Australian ISP’s would have a greater chance of delivering a chunk to an Australian client, than an Asian farmer. Does that hold water you think?
Depends on the size of the farm. You are looking at the proportion of nodes in the farm to the rest of the world. If the farm has 10,000 nodes and the world has 1,500,000 nodes then it will have a little benefit in latency some of the time, but doubt it would be noticed by anyone.
Look at simple probability, if Asia has 1,000,000 and the world has 1,500,000 nodes then 1 in 150 chance any one hop occurs to your farm and 2 in 3 chance it occurs to Asia.
Yes, I didn’t consider that, but I also know that China is not all of Asia, they have only a third or something like that the population of Asia. India is actually similar population to China and has a lot of PCs too. and of course Japan, and all the Asian nations dotted over the region.
Personally I doubt that Asia will be the dominate force in SAFE for a while. Its hard to know though, and of course the GFW slows down a significant portion of the Asian Nodes.
) it’s around 1 GB. And when I’m done watching and check uTorrent I’ve uploaded about 5 times that amount before I close the torrent. That’s how SAFE should work IMO. Just running in the background providing stuff to the network. And when 1 user is quite fast it will support a user in India for example who has a very slow connection.