Slashdot Mirror
Canadian Telco Telus Moves All Call Traffic to the Net
justice0x1 writes "An article on the Tornonto Star about Canada's Telus making a large scale motion to move all call trafic over to IP caught my eye today. 'Telus will become the first dominant phone carrier in North America to make the risky transition, a move much talked about and which Telus will make happen on a dramatic scale.' Since I work in the Telus Internet Service department, it will be interesting to see exactly how this new technology fares. Seems almost premature to me, but I guess it's all or nothing with telecomunications these days; you need to get an edge on the competition somehow. Why not start by moving youre entire long distance network over to IP?"
Just because it's VoIP, it doesn't mean it has to flow over the "public" IP network. They can simply point the VoIP over their own network and still see massive savings.
For the VoIP/PSTN imparied, VoIP traffic can be optimized (that's compressed) down to 8K (or more, but a cost of clarity) as opposed to 56K for traditional circuit switched telephony traffic.
And yes I have worked with the technology.
I use Vonage VOIP for my home phone and they require 90kbps both ways in their default config. They have an option to reduce the call quality, and thus the size. 90kbs is about 2/3rds the way to standard phone quality comming from cell phone quality, if that makes since to you.
but what do i know, i'm just a model.
In europe, traditional circuit-switched telephone networks use channelised E1s, which basically have 15 64kbit/s data channels, along with a 64kbit/s control channel, giving a 2Mbit/s channel.
Using H.323 (the main VoIP standard) you can choose from a number of compression codecs, from 32kbits upwards, although there exists a couple of codecs which sound pretty good (coparable to a average analogue phone line) at 64kbit/s, to you can make the transition without effecting you capacity.
I was involved in some trials here in the UK, and (IIRC) we were using about 90kbits, and that was _crystal clear_ - the thing you most notice is that when no-one is talking, it is actually *silent* - you don't realise it until you try VoIP howe much noise is in a normal analog call.
This is fact one of the advantages of VoIP, you can have in effect variable bitrate encoding, as if no-one is talking, then you don't have any data to encode!
" To steal ideas from one person is plagiarism; to steal from many is research. "
The main VoIP standard is H.323 - Check out OpenH323, an open-source implentation of this technology.
" To steal ideas from one person is plagiarism; to steal from many is research. "
..I use Vonage for my phone at home. I do not even have a land-line installed in my apartment. I find the features to blow any phone company away, and the price and service is excellent.
Besides the overall geekness of being totally VoiP, I have had nothing but good experiences with it.
And I get to have an LA and NY number....
Rob
Nope!
E1 is 30 B channels at 64kbits/s + 1 D channel at 64kbits/s for signalisation plus 1 synchronisation and misc operator stuff at 64kbits/s. So E1 is 32*64k=2048Mbits/s.
Some operators have services for 10, 12, 15, 20 or 24 channels depending of the offert but this is simply a limit of concurrent B channels open in a 30 B channles capacity E1 line.
As a few posts have already descibed, VoIP is not the same as voice over the internet. The VoIP that Telus is using is actually being carried over dedicated data circuits which offer, not only high speed (OC-12 +), but also end to end Quality of Service control. Indeed, these dedicated data circuits may well carry some internet traffic as well, but Telus is able to run the voice traffic at a higher priority than the less time sensitive data.
In fact, this is already being done by several carriers including Sprint, MCI, Intermedia, Verizon and probably others.
I have also installed numerous private networks utilizing Nortel or Cisco equipment to carry VoIP over dedicated private networks, usually frame-relay WANs. These VoIP calls are 100% reliable and are perfectly clear.
In two cases standard internet connections (cable, xDSL, frame-relay) were used to carry calls between several different offices in the United States, Canada, Europe and Mexico. These connections are not always as clear as those running on private WANs but, they have proven to be 98% reliable and are indistinguishable from regular land lines, in terms of clarity, 85% of the time.
The article is not technical, and this is pure market speak. What they were adding was probably "comfort noise." Comfort noise is placed in the dead air where you aren't actually sending sound (to save bandwidth).
Another thing normally added is near-end echo. This is the sound of your own voice piped back at you with reduced volume, something that happens in the standard phone system naturally. People have become so used to hearing the echo that they think something is wrong if they don't get it. Because of this, you see a lot of people yelling at their cell phones.
Try learning more modern stuff than whatever antiquated VoIP crap your company was using.
As an example, I was working with VoIP stuff 3 years ago with Cisco, T-Mobile and Level 3: we had a project that took a GSM signal, put it onto a local (in building only) network, and from there to VoIP and onto the Cisco Call Manager system. This allowed the cell phone to be your internal company phone extension while on campus, but automatically switch back when off campus. We faced many of the same problems you cited but all were solved by the end of the "beta" test.
Also we deployed a nation-wide softswitch infrastraucture that allowed us to hook Cisco VoIP phones at any point in the network and make toll-quality calls out from the main PBX back in Broomfield. This saved the company from having to drag a copper pair to each and every little regeneration site onthe fiber right of way, which is not an inconsiderable amount of cash when you figure the extent of Level 3's network.
Update your knowledge base. IP is starting to be picked up now that the private IP networks Like Level 3, Genuity, plus parts of AT&T and Sprint have worked out the problems. Draft Martini (Read the IETF documents) has been delivering standard phone services like ATM and FR over IP for a year. And Level 3, where I used to work, has had an all IP infrastructure for several years now, using the old XCom (Now VIPER) sofswitch. Want to know where those cheap long distance calling cards are coming from? They probably travel over Level 3's all IP network, converting at the edges on the softswitched architecture. And a good percentage of Worldcom/MCI long distance moves on Level 3's all IP network as well.
Secondly, the VoIP mentioned here is primarily in the backhaul, behind the CO switches (RTFA again!). For Inter and Intra-LATA carriage, VoIP is highly reliable, and much less expensive than the TDM stuff.
As for the "silence", RTFA. Seriously, they solved this one a long time ago with a little echo back on the reciever side, as well as "comfort noise" from the IP switch. This technique goes back several years.
Seriously, you need to learn more - and RTFA, because you are "mistaken about a great many thing"....
Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo! http://goo.gl/J9bkO
Read my subject again, you know what "!=" means right?
The article *title* "Canadian Telco Telus Moves All Call Traffic to the Net" implies that Telus will be using the public Internet.
The *article* says several things but doesn't clearly state for non-technical readers that the network that Telus will be using for this VoIP service is their *private* network. The article talks about the poor quality of previous Internet phone call implementations without qualifying that assertion with something like "but Telus will use their own network instead of the Internet."
Others have pointed out the same thing. The title is misleading.
If you don't want to repeat the past, stop living in it.
Indeed, the title is misleading. The traffic will flow over TELUS's national Private IP network which, incidentally, happens to be the fastest optical/IP network in north america end-to-end.
----------------------------
Esobofh - Currently drinking fresh mango juice.
Try entering the term in Google, I got 540,000 hits.
Actually the sample rate is 56kbit/sec on voice calls, which is where the 56k maximum theoretical speed of a modem comes from. 64kbit/sec is a full B channel. In the old days everything telco related was based around B channels (A T1 is 24 B channels for example). Most of the "old" telcos still around are still based on this. The newer companies have to work with it to interoperate but internally their networks probably aren't based on it.
There's a big difference between desirable internet traffic flow and desirable voice traffic flow though. Voice traffic is sensitive to latency and changes in latency far beyond most internet traffic. Voice phone calls are supposed to be real-time two way conversations, if data arrives "late" in a phone call you may as well just discard it.
Your minimum overhead is going to be something like (Layer 2 header per pkt + IP Header per pkt + higher protocol headers per pkt) / ( 56kbit/sec * (maximum acceptable one-way latency) ) so if you're willing to accept > 1 second worth of latency in each direction then you can get that down to about 1% overhead, if you're aiming for 1/100th of a second one-way latency you'll have about 85% overhead. These are just minimum latency values attributed to buffering a certain amount of data before sending it - as required for a packet based network - nothing says upstream or lower layer handling wont add more on top of that (and since it takes some finite amount of time to transmit the data it surely will, but hopefully not anything noticable).
IP isn't designed for this kind of thing at all though. Surely it can carry any data you want it to, but when you're talking about wanting to send data in 30 byte chunks to minimize the latency between the first sample encoding at the source end and decoding at the far end, the IP header is going to contain a lot more information than is practical.
But hey, if you have piles of money it doesn't matter if you go about things ass-backwards. Just type in "osprey crashes" to google for a good military example!
So I suppose you still can not slashdot the phone network..
Actually you CAN slashdot the PSTN (Public Switched Telephone Netowork). And you always could. The equipment is sized to handle the expected peak loads with some slop. But there is nowhere near enough equipment in place to handle every phone being connected to another phone.
You can slashdot it at several levels. The commonest is the "all-trunks busy" level - where all the routes from the calling phone to the called number (that the switching equipment knows how to use) are busy. In the older exchanges that produced the tones that sounded like a busy signal but twice as fast. Modern stuff gives you a recording.
You can also tie up all the equipment that gives you a dial tone and collects the digits you dial, by getting enough people on the exchange to try to make calls at once. Usually this just means you wait a second or so for a dial tone - and maybe not even notice it. If it's REALLY severe you might wait seconds, or minutes, and then it is really noticible. But it's also really rare.
The last time I recall that actually happening where I lived was the Loma Prieta earthquake, and the time before was the assassination of JFK. Before that was at an old relay-based exchange (using line-finders rather than registers) where the line finders didn't time out, and a tornado had shorted out enough lines - which made them look "off-hook" - to busy out all the line-finders that could give my phone a dial tone.
Again, modern equipment is more informative: When things get hairy the people operating the network can switch it to a mode where, when you take your phone off the hook, it first connects you to a recording asking you to hold off unless it's an emergency, then giving you a delay followed by a fair chance at a dialtone. (I THINK it actually deliberately delays you a bit even if it COULD have given you a dial tone right away, both to throttle you and to give you a chance to hang up if it wasn't urgent.)
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way