Is end user VOIP viable without large percentage of BB users?

watty

Discuss
http://www.theregister.co.uk/2006/03/23/wanadoo_voip/

Wanadoo UK chalks up 150K VoIP users

In a statement the France Telecom-owned ISP, which used to be known as Freeserve but is morphing soon to Orange, said: "The news confirms Wanadoo's position as Britain's Number One phone to phone consumer Voice over Internet Protocol (VoIP) provider."

For the record, a BT spokesman said the firm had around 100,000 users of its BT Broadband Voice VoIP service.

According to the latest stats from industry regulator Ofcom, there are more than 500,000 active VoIP users in the UK benefiting from "a significantly lower-cost alternative to traditional fixed-line calls". And since more than 10 million households in the UK now have broadband, this number is predicted to grow rapidly.

Two weeks ago it emerged that France Telecom had chalked up more than a million residential VoIP accounts across Europe, with some 150,000 new subscribers being added each month across France, UK, The Netherlands, Italy and Spain

crawler

i think Watty should set secondary school state exams....

"blah blah blah....." discuss

Blaster99

Do people like BT and FT apply the same quality to VoIP as they do to POTS calls? I would guess not. Must be a difficult one for them to market as people have very high expectations of a telco.

Tom Young

Watty,

End user VoIP can be origination of calls or termination of calls. There are so many differing types of Voice as an application of the IP space that quality would be a concern and of course a variant.

Direct answer: YES to, 'is it viable without large percentage of BB users?'

Take your pick though on the platform and methodology of choice.

Quaere? Define BB please

watty

Basic useful BB definition

120K bps or faster, upload or download.
Always on
FIXED price, irrespective of time, or data transfered.

If there is a cap (and there are arguments both ways) any transfer cap should be large enough that it is of no concern to ordinary users. Say at least 5G per month per Mbps of max bandwidth (i.e. 4Mbits = 20G cap).

Tom Young

Watty,

Great. That's what I had in mind. Now what exactly are you stroking at in the question on viability?

Reason for asking: An IP connection with a SIP Proxy available to serve a user might originate calls and receive over IP, but be converting them to TDM - Time Division Multiplexing along the way. For that again we may elect to chat on a Peer-to-Peer network, e.g. Skype, which I note you are using on your profile.

Do you have specific criteria or a criterion you want to exhaust on viability?

There are plenty or problematic technologies out there that in my view are not viable for a few reasons e.g. SBCs or Session Border Controllers.

Just a view, I want to know what exactly you are after and hopefully provide an adequate answer.

Tom.

watty

I think that without a certain critical mass of BB among friends, VOIP is of no interest. Other than transcontinental friends/family it has no interest to the majority of dialup people. Unless someone has an aways on connection with no per time or per data charge, VOIP is unattractive.

When a fair percentage of friend have BB and a couple get Skype then they all will.

Interms of VOIP as way to connect to people with real phone lines and real mobiles there has to be equipment working independant of PC. If you have to pay Eircom line rental anyway, VOIP may be irrelevent for local calls to non VOIP users.

Not all BB is equal either. Most dialup connections seem better for Skype than a typical Ripwave IBB BB connection.

I'm not "after" anything. In terms of user if they can dial out and recieve calls from PSTN/Mobiles and the voice is intelligable and connection not dropped (QOS issues) they don't care how it works.

But it seems to me that VOIP is only of interest to actual end users as a particular technology if broadband is generally available.

Users don't really care about technology. They want cheap reliable voice calls and text messaging. FDM, TDM, POTS vs ISDN Local loop, VOIP etc is not in that sense relevent.

Blaster99

Broadband is not exactly an unusual thing in the big wide world. It's only really an issue in Ireland and a few other countries with third world infrastructure.

One of the issues technically, as I highlighted on the VoIP forum, is that the VoIP codecs like G.711 and G.729 are designed for TDM, not packet networks. Skype on the other hand uses wideband codecs specifically designed for packet networks that can handle stuff like packet loss a lot better and in my opinion Skype-to-Skype generally works much better than SIP/RTP/G.7xx on a similar connection.

There are still VoIP applications over ADSL. I have a company VoIP phone at home. When somebody rings it, it rings at my desk in work and at home and I can pick up either. I make any business calls on my IP phone, regardless of location. I have a soft phone on my laptop so when I'm out and about I can make calls from it too. The seemless roaming aspect of VoIP is v cool. So long as the network connections are loss free and latency is reasonably low and stable. This is of course not always the case so quality can suffer. I think VoIP over the plain ol' internet has a bit to go yet.

If you have a need to have multiple phones you can also use VoIP over ADSL very cost effectively. If the regulator had anything resembling balls they would do like in Sweden where you can get a data-only ADSL line with a significant reduction in line rental. If that were to happen here, VoIP would be a serious competitor to ordinary voice and it would also drive BB uptake I would think.

useruser

Blaster99 wrote:

One of the issues technically, as I highlighted on the VoIP forum, is that the VoIP codecs like G.711 and G.729 are designed for TDM, not packet networks. Skype on the other hand uses wideband codecs specifically designed for packet networks that can handle stuff like packet loss a lot better and in my opinion Skype-to-Skype generally works much better than SIP/RTP/G.7xx on a similar connection.

Skype reportedly are using the ILBC CODEC which is better in lossy conditions. Most broadband connections do not have packet loss problems however so VoIP services will work just fine with G729 or G711 (depending on available bit rate). If you have a lossy connection ILBC will sound better. Otherwise, G711 is toll quality, G729 is just as good as ILBC with low/no packet loss. If QoS is implemented then ILBC has no advantage.

watty

useruser wrote:

Skype reportedly are using the ILBC CODEC which is better in lossy conditions. Most broadband connections do not have packet loss problems however so VoIP services will work just fine with G729 or G711 (depending on available bit rate). If you have a lossy connection ILBC will sound better. Otherwise, G711 is toll quality, G729 is just as good as ILBC with low/no packet loss. If QoS is implemented then ILBC has no advantage.

Which IMO is why Skype can work well on a 33K dialup.

Anyhow NETWORK use of VOIP is not really an issue. The original title was asking about END USER use.

Does desire to have VOIP really increase a users dxesire to have BB? Some reports have claimed this. Personally I think VOIP for end user is an additional advantage of BB after you decided and managed to get it, not an app driving BB demand.

Is VOIP "icing on the cake" for BB users or really driving demand for BB?

If it is more "icing" than driving demand (maybe a bit of both), then my opiniion that you need a critical mass of BB users in a geographic area for ENDUSER voip to "take off" is true.

Blaster99

Skype uses iSAC, by the way, which is a VBR wideband codec that works down to 16kbps. There is plenty of packet loss out there, by the way. Architecturally it's insane to use codecs designed to deal with bit loss on packet networks.

VoIP means cheap (or in many cases free) phone calls to POTS users today. And the momentum is already there seeing as there are lots of providers and users.

useruser

Blaster99 wrote:

Skype uses iSAC, by the way, which is a VBR wideband codec that works down to 16kbps. There is plenty of packet loss out there, by the way. Architecturally it's insane to use codecs designed to deal with bit loss on packet networks.

VoIP means cheap (or in many cases free) phone calls to POTS users today. And the momentum is already there seeing as there are lots of providers and users.

We're both right - Skype uses iLBC and iSAC. Why do you think "it's insane to you use codecs designed to deal with bit loss?"

Blaster99

I think iLBC has been deprecated actually. It offers no benefits over iSAC, only negatives. iLBC requires 13kbps (from memory) so it's no more narrow than iSAC and can't do wideband where available.

With regard to bit loss on packet networks, I hope you are taking the piss.

watty

AFAIK X.25 (packet network) provides error free connection at user level, as does ISDN/TDM. Both for a given connection have fixed latency.

IP networks do not, nor do they have fixed latency. I have seen Latency on IBB connection vary from 200ms to 1750ms during a call. But all TCP/IP networks have non-determinate routing and thus potential variable non-predicable latency. Using UDP, packets are just lost. Using TCP, packets are resent at expense of raising latency AND reducing bandwidth.

So for G'teed QOS limits X.25 and TDM (ISDN etc) can have pre-known bandwidth, known latency and no packet loss.

For IP networks even during call, the bandwidth and latency can vary and is always unknown in advance. A given QOS can't be g'teed. You can have error free, but worse latency/bandwidth QOS with TCP, or unknown error rate (packet loss) and more stable latency and bandwidth (but still unknown in advance and unstable) using UDP.

This is the classic trade off between a connection orientated network (Traditional PSTN / X.25 and ISDN/TDM) and a non-deterministic redundant packet forwarding network with no fixed connection/route per call.

TCP/IP internetworking was not designed at all with time sensitive content like voice and video. It's actually quite an achievement it does it as well as it often does. But statistically, no g'tee.

watty

I can't see how anyone can offer a known good minimum QOS for VOIP on 100% TCP/IP network.

useruser

Blaster99 wrote:

I think iLBC has been deprecated actually. It offers no benefits over iSAC, only negatives. iLBC requires 13kbps (from memory) so it's no more narrow than iSAC and can't do wideband where available.

Last couple of calls I made were iLBC - any idea how it chooses? (v2.0.0.76)

With regard to bit loss on packet networks, I hope you are taking the piss.

Not really, I get your point about bit loss but didn't you mean to say "inefficient" rather than "insane?" I assume you would accept that G729a is the most popular low complexity, low bandwidth codec currently in use for VoIP? (not including Skype obviously!)

useruser

watty wrote:

I can't see how anyone can offer a known good minimum QOS for VOIP on 100% TCP/IP network.

It's pretty easy - Diffserv, WFQ/LLQ etc. what's so hard about that?

watty

Since the Internet is designed to to survive entire cities getting removed by thermonucular devices and no explicit route can be g'teed, I'd have to disagree. The Internet is designed to have only one level of QOS, Delivery. Time is not possible to specify.

In a headlong rush to abandon circuit switched topologies etc and embrace TCP/IP by VOIP operators this tends to be forgotten.

However VOIP is usefull, is often better than GSM, but rarely as good as ISDN end to end. Ironically while Germany uses ISDN as standard and can run DSL over ISDN, broadband here involves removal of ISDN and stepping back 80 years in phone technology.

Blaster99

ISDN doesn't magically make the phone line better. It's still the same copper. ISDN was an evolutionary dead end that's unusually popular in Germany. They have lots of nice true flat rate packages there, though.

useruser is referring to networks that are managed to run VoIP, in corporates and internally in telcos. In those situations you can obviously guarantee quality to a great degree and VoIP tends to work flawlessly. In the greater unknown, that's not the case. Nobody is really going to prioritise on DiffServ etc as everyone will just mark their traffic as high priority.

G.729 is very popular but my point remains, and it's remarkable that the VoIP world keeps plugging on with stuff that's patently unsuitable for the task. I suppose G.729 chips are cheap [as chips] these days.

useruser

When you say "VoIP," do you mean "Voice over the public Internet" or the technology as a whole?

watty wrote:

Since the Internet is designed to to survive entire cities getting removed by thermonucular devices and no explicit route can be g'teed, I'd have to disagree. The Internet is designed to have only one level of QOS, Delivery. Time is not possible to specify.

Hmmm., that may have been true of IP networks 10 years ago but most MPLS cores can & do guarantee explicit routing. There are far far more VoIP minutes traversing private networks and interconnects than crossing the public internet but with inter-provider MPLS there is no reason that QoS can not be guaranteed end-to-end if required.

However VOIP is usefull, is often better than GSM, but rarely as good as ISDN end to end. Ironically while Germany uses ISDN as standard and can run DSL over ISDN, broadband here involves removal of ISDN and stepping back 80 years in phone technology.

NGN and IMS architectures are entirely VoIP, all voice traffic will be VoIP at some stage as operators move to these. Any half decent broadband connection will easily provide an ISDN quality VoIP call with g711 - the bandwidth required is tiny, 110kbps tops. The wideband codecs will provide an even better quality call than ISDN with lower bandwidth requirements than g711 (but higher cpu cost).

useruser

Blaster99 wrote:

G.729 is very popular but my point remains, and it's remarkable that the VoIP world keeps plugging on with stuff that's patently unsuitable for the task. I suppose G.729 chips are cheap [as chips] these days.

I assume that is the case (price of DSPs), however I see that some ATA manufacturers are beginning to support iLBC (and I think Cisco announced that it will support it in their gateways too).

I still don't quite get your point though, g729a works very well and the bandwidth savings over g711 are obvious (not such a big deal for single lines but for business users it's important) - patently suitable for the task I would say! What's your objection?

Blaster99

Are we running around in circles here by any chance? The internet is packet lossy. G729 (and G.711) can't deal with that. Codecs used on inherently packet lossy networks should obviously be optimised for that. To keep plugging away with stuff designed to deal with non-existent bit loss is pretty dumb. In my world it is, at least. But how and ever, I have said what I need to say about this and I'm just repeating myself over and over.

I'm toying with setting up an Asterisk G.711 -> iLBC transcoder but I've kinda put it on the backburner because I don't really have any packet loss to test it with.

watty

Yes well ANY codec using IP format on a fibre will be fine. But for Public internet, VOIP END USERS, not specially designed networks and carriers, QOS is a problem.

However the public will use even PC based VOIP more and more when a critical mass of BB users exist.

Phone Networks that use VOIP is a different beast.

Blaster I can give you an email of someone on IBB Ripwave in Limerick... If they agree. Lots of stuff to stress a system

watty

I think some of us have been talking about quite different things.

useruser

Blaster99 wrote:

Are we running around in circles here by any chance? The internet is packet lossy. G729 (and G.711) can't deal with that. Codecs used on inherently packet lossy networks should obviously be optimised for that. To keep plugging away with stuff designed to deal with non-existent bit loss is pretty dumb. In my world it is, at least. But how and ever, I have said what I need to say about this and I'm just repeating myself over and over.

I see what you're saying, trouble is that the codecs that can handle packet loss require a lot more processing power. Even g729 requires twice the dsp resources of g729a and that can mean 1000s of $ "wasted" in Cisco gateways for a minimal improvement in MOS. Out of interest, do you happen to know what codec skype uses for PSTN calls? Do their PSTN gateways support ILBC and iSAC?

I don't regard the internet as inherently lossy though - don't get me wrong, I acknowledge that there is congestion and loss but it's just not that bad in my experience. The fact is that these codecs will work fine (and lots of voip operators are proving this every day) across most broadband links without QoS.

Blaster99

It's a funny thing that there's so much talk about DSP implementation of codecs etc but a lot of VoIP service providers use Asterisk which uses software codecs. I haven't studied the service provider area all that much but it seems to me that these guys probably don't run tests to ensure what the call capacity of their Asterisk nodes are if you have lots of CPU expensive G.729 calls in progress, because from my own experience when it comes to developing PBX's that kit is mega expensive. I would also hazard a guess that they try to bridge the codec to the other side as much as possible to avoid expensive transcoding, so if it's G.729 inbound they negotiate G.729 outbound to the backhaul too. I incidently suspect this is why G.729 often deliver much better results than G.711 despite my connection to the service provider is practically perfect so the codec should make no difference on the "local loop". While neither codec can deal with packet loss, the smaller frames with G.729 leads to less latency, less opportunity for packet loss, and often removes echo from calls.

Sorry, I don't really know all that much about Skype. Problem is that they're telling nothing but lots of people have made educated guesses based on tests. SkypeOut sounds pretty bad usually, bad as in like a tin can.

useruser

Tom Young wrote:

There are plenty or problematic technologies out there that in my view are not viable for a few reasons e.g. SBCs or Session Border Controllers.
Tom.

Can you expand on this point Tom?

useruser

Blaster99 wrote:

It's a funny thing that there's so much talk about DSP implementation of codecs etc but a lot of VoIP service providers use Asterisk which uses software codecs. I haven't studied the service provider area all that much but it seems to me that these guys probably don't run tests to ensure what the call capacity of their Asterisk nodes are if you have lots of CPU expensive G.729 calls in progress, because from my own experience when it comes to developing PBX's that kit is mega expensive. I would also hazard a guess that they try to bridge the codec to the other side as much as possible to avoid expensive transcoding, so if it's G.729 inbound they negotiate G.729 outbound to the backhaul too. I incidently suspect this is why G.729 often deliver much better results than G.711 despite my connection to the service provider is practically perfect so the codec should make no difference on the "local loop". While neither codec can deal with packet loss, the smaller frames with G.729 leads to less latency, less opportunity for packet loss, and often removes echo from calls.

I don't imagine many service providers are using Asterisk as the PSTN gateway? (as the softswitch sure, but if you are doing any volume of minutes at all then it just makes more sense to go with dedicated hardware).

Interesting comments on quality, transcoding can certainly have a noticable detrimental impact but I am surprised that you are hearing much better results with g729 than g711 even so. Service providers will not generally transcode - RTP just gets handed over as you describe.

Sorry, I don't really know all that much about Skype. Problem is that they're telling nothing but lots of people have made educated guesses based on tests. SkypeOut sounds pretty bad usually, bad as in like a tin can.

Disappointing that it should be so poor, I'll have to set up a SkypeOut account and try it out.

Blaster99

I think SkypeOut is possibly deceptively poor because Skype-to-Skype is usually wideband and sounds excellent. But also I never really got the impression that Skype cares all that much about SkypeOut.

As I said, I don't know how VoIP providers operate so I'm just speculating. My assumption is that most of them are buying PSTN access from other companies in different parts of the world so they're using SIP or IAX to somebody else, so they're just bridging calls between VoIP endpoints and collecting money in the middle. I guess a premise of VoIP is that you terminate the VoIP call in the country of destination and charge the equivalent of a local call as the international transit cost is essentially the same to all destinations. But that's just how I imagine that it works, I've never looked into it with any detail. I have noticed significant difference in call quality to different countries with everything being the same at my end.

I've used the VoIP service provided by an Irish PATS as well and I've found them to be the best but I haven't really extensively tested with them. They're not using Asterisk from what I recall from SIP traces and being a PATS I would imagine that they have more reliable interconnect arrangements for call termination. Just idle speculation on a Friday night, though. I'm not mentioning any names for that reason.