VOIP = square peg, round hole ?

by John Q on September 24, 2004

The failure of yet another VOIP (voice over Internet Protocol) company reminds me of a question to which I’ve never got a satisfactory answer. Is there any technological reason for thinking that VOIP is a good idea? To clarify, IP is a packet-switching technology, which is great for data, but everything I’ve read suggests that circuit-switching is better for voice (that’s what it was designed for, after all).

Whenever I’ve seen an explanation of the supposed advantages of VOIP, it has seemed to involve regulatory arbitrage. That is the technology is supposed to use cheap substitutes for regulated voice lines, while typically relying on those same lines for access either via old-style modem or ADSL.

But all of this is just an impression. Can anyone set me straight?

{ 27 comments }

1

Jason McCullough 09.24.04 at 1:44 am

Circuit-switching is just packet-switching with bandwidth and delay times in dedicated allocation for the lifetime of the call. It costs more (inefficiences in predicting allocation, re-routing is harder, some other boring theoretical maximum efficiency results), but is more reliable.

Voice is chiefly a delay-bounded activity (can’t exceed X milliseconds or it will suck); data isn’t. Hence circuits working better for voice, as to the lack of widely-adopted delay-bounding infrastructure on the data networks.

This is outweighed by the ludicrous cost of VOIP on regulatory barriers and monopolist tolls, though, or it looks like it *should* be.

2

Jason McCullough 09.24.04 at 1:48 am

Circuit-switching is just packet-switching with bandwidth and delay times in dedicated allocation for the lifetime of the call. It costs more (inefficiences in predicting allocation, re-routing is harder, some other boring theoretical maximum efficiency results), but is more reliable.

Voice is chiefly a delay-bounded activity (can’t exceed X milliseconds or it will suck); data isn’t. Hence circuits working better for voice, as to the lack of widely-adopted delay-bounding infrastructure on the data networks.

This is outweighed by the ludicrous cost of VOIP on regulatory barriers and monopolist tolls, though, or it looks like it *should* be.

3

Matt 09.24.04 at 1:49 am

This doesn’t address your question at all, but there was an article in the U Penn student paper today saying that Penn is planning to switch to VOIP for all of its dorm room phones (maybe just for internal calls- I didn’t read the article that closely) w/in a few years. We’ll see what happens.

4

PG 09.24.04 at 1:58 am

The only time I’ve used VOIP is on 9/11/01, when no one’s cell phone seemed to be working but the internet was humming along fine. It assured me that my parents were safe, and that was Good Enough for Me.

Also, aren’t politicians proposing massive subsidies for broadband? This could make VOIP a more sensible proposition than it is in the current internet landscape.

5

michael 09.24.04 at 2:07 am

VOIP makes more and more sense as bandwidth and QoS improve. If you can keep low latency and ride on top of a global network for virtually nothing, then do it.

It makes a lot of sense of business use – rather than having to worry about having your technicians or telco provider maintain your phone systems, employees can simply pick up their existing eqipment and move it from room to room. No cross-connects or anything.

6

michael 09.24.04 at 2:10 am

pg: why are we going to subsidize broadband? I haven’t heard anything about this…

7

Brian 09.24.04 at 2:25 am

Clearly VoIP is a far more effective solution than traditional circuit switched technology. If you look at the major carriers in the world they are moving their core networks to VoIP based systems. So, whether you know it or not, and whether the start and end of your call may be circuit switched, the majority of most calls within the next couple years will travel over packet switched networks.
MPLS is the prime core network switching system which is far cheaper to run than traditional ATM or Frame Relay neworks. If you are a carrier this is what you are doing to reduce your costs.

As to why another VoIP provider folded, that’s another issue. As with any new technology there are going to be companies that find their way and ones that get lost along the route. AT&T’s CallVantage, Vonage, and others are showing the great features you can access with VoIP – things like nomadic mobility, the death of long distance. Remember, in a packet network, packets don’t care how far they travel to get to the end. (You don’t pay extra for accessing a website in New Zealand) The same goes for VoIP. You are paying for the bandwidth, not how far the electrons travelled to get to you.

Bell Canada has committed to have its network transferred to IP by 2006. That’s soon, and it’s a large network covering most of Canada. That’s not a whim, that’s business reality.

8

Jason 09.24.04 at 2:33 am

I think a big part of the issue here is legacy pricing. This is backed up by the fact that voice isn’t very much data (in comparison to big files), and so long as we have excess bandwidth, using voice with the slack is no problem.

I helped teach a pricing in the telecommunications industry class a couple of years ago, and there is a price to terminate a call on someone else’s exchange (eg. 10c/min). This normally comes into play in international calls. It is symmetric between exchanges, but by virtue of economic differences, there are a large number of calls out of rich countries and into poor ones. Thus the poor countries get a decent amount of revenue from this.

One notable example is Dutch Guinea, where the average phone time in the country is more than 24 hours per person per day. The reason is that there are a number of 1-900 numbers (pay per minute) that are set up so that when you call them, they route you to Guinea, and then have another call from the US to Guinea. Connect them in Guinea and it comes up on your phone call.

9

Jake McGuire 09.24.04 at 2:58 am

Circuit switching is better than packet switching for voice traffic, everything else being equal. But everything else isn’t equal – on the one hand, the demand for data communications is driving the improvement of packet switching technology much harder than circuit switching technology. On the other hand, computers are REALLY FAST these days, so brute-force methods – like figuring out how to route each individual packet of a phone conversation instead of figuring it out once at the start of the call – become practical.

10

c 09.24.04 at 4:07 am

Dutch Guinea? Do you mean Surinam

11

Phill 09.24.04 at 4:07 am

It is an optimization issue. At this point the amount of voice data being moved is a small fraction of the rest. It is cheaper to use one network for both and accept a small amount of inefficiency.

The other issue is that while the telcos recovered from their capital spending binge in the 90s they stopped buying any new stuff. So they litterally have no suppliers. Lucent Marconi and Nortel are all pale shaddows of their former selves, they have zero research dollars. Cisco and IP switching are the only games left in town.

The other change is that with quality of service parameters it is now possible to reserve bandwidth in advance.

12

rps 09.24.04 at 4:07 am

Asking whether packet switching is better than circuit switching is asking the wrong question. If packet switching networks are good enough to carry on a conversation, what difference does it make? I knew someone who has VOIP and he said it was fine.

Also, don’t read too much into some companies going under. Twenty years ago you could have pointed to any number of failed computer-related companies; those companies are forgotten and computers are now ubiquitous despite their failure.

13

rps 09.24.04 at 4:09 am

Asking whether packet switching is better than circuit switching is asking the wrong question. If packet switching networks are good enough to carry on a conversation, what difference does it make? I knew someone who has VOIP and he said it was fine.

Also, don’t read too much into some companies going under. Twenty years ago you could have pointed to any number of failed computer-related companies; those companies are forgotten and computers are now ubiquitous despite their failure.

14

Marcelo Rinesi 09.24.04 at 4:28 am

There is also the factor of, for lack of a better term, “convergence”. With VOIP, I can do things with voice calls that I couldn’t with the legacy network, at least not without specialized, expensive, rigid hardware. Weird filtering and forwarding, data attachments, noise cancellation, whatever you can come up with. Building things in software always results in accelerated improvement curves.

15

rps 09.24.04 at 4:46 am

Asking whether packet switching is better than circuit switching is asking the wrong question. If packet switching networks are good enough to carry on a conversation, what difference does it make? I knew someone who has VOIP and he said it was fine.

Also, don’t read too much into some companies going under. Twenty years ago you could have pointed to any number of failed computer-related companies; those companies are forgotten and computers are now ubiquitous despite their failure.

16

Neil Katin 09.24.04 at 5:23 am

One other difference between the VoIP/TCP/IP data networks and the circuit switched telephony network: in the current internet application intelligence tends to be at the “edge” of the network. In the telephony networks it tends to be embedded in the fabric itself.

This makes it much easier to deploy new services. No router had to be changed to support the world wide web; it just worked over the existing network.

VoIP worked the same way: it could be a new application that piggybacked on the existing infrastructure. It is for this reason (more than for any regulatory arbitrage) that it is so affordable – it does not take a dedicated infrastructure.

17

me2i81 09.24.04 at 6:00 am

The biggest issue is, everyone needs a data network these days, and in reality everyone needs a reasonable “ping” time, so everyone’s (non-dialup) data network is capable of carrying voice. Once you’ve built out a data network, adding voice is “free,” since it uses relatively little bandwidth. Also, with digital signals, voice is no longer a fixed bandwidth because of compression. As a result, as Jason pointed out, in reality there haven’t been actual circuit-switched networks for a long time–what we have had for years is a somewhat anachronistic system: copper loops going to a “switch” which digitizes the signal and multiplexes it using packet switching. VOIP is just recognizing that there’s no reason to run analog on the final loop if you don’t have to, and recognizing that there’s no longer a need for the proprietary packet bus inside the switch when you can buy gigabit IP switch technology off the shelf for almost nothing, so you might as well use IP. The latter is already happening without anyone noticing.

VOIP has been most successful in large corporations. The one I work for adopted the Cisco system, because every desktop already has a reasonably high-bandwidth net connection, so why not dump the PBX and all of the trunk lines between offices? It saves a lot of money. Once everyone is used to that at work, VOIP at home will have a better market acceptance, as long as it doesn’t get screwed up by thinking it’s something other than data, which requires that we understand that all our POTS telecom regulations have become obsolete. I know, good luck…

Consumer VOIP services probably fail because some of them suck and their pricing model is often ridiculous.

18

Randolph Fritz 09.24.04 at 6:32 am

VOIP really shines in corporate networks, where it saves pulling a lot of cable and the installation of expensive switching equipment; the company data net carries the traffic for a very modest charge. In a heavily wired office that is a substantial savings.

As a residential service, it is still finding its way. It does provide a a very high level of telephone service without the nickel-and-diming charges for voicemail, call waiting, caller id, and so on, and on–all that can be done in a VOIP station (telephone or computer) for free. There is the potential for higher-bandwidth audio, providing higher-quality sound. And, again, there is a savings in hardware, though this is less significant for a residential user. In the longer run, it could be used as a multi-media components, so that one can transmit images, even video, alongside a telephone conversation. There is also the possiblity of integration with other sorts of network services–IM and e-mail, for instance.

That said, service quality is going to be poor until the people who own the cables, usually the RBOCs, provide limited-latency bandwidth reservation at modest prices, which isn’t likely anytime soon, since they make money by charging for voice and data service separately.

19

Maynard Handley 09.24.04 at 7:51 am

John, at least in the US, immigrants are using VoIP everyday. Go to a website like justphonecards.com and you can buy phone cards that allow you to talk to places like China for around 3.5c a minute. Those are based on VoIP for the bulk of the distance, with just some weirdness at both ends to hook into the phone system.

I’ve used them frequently and they work just fine — every so often you get a bad connection, but no more or less often than you get on analog connections. This, and the nature of the bad connections leads me to believe that the problems in both cases are in the analog bits at both ends, not in the VoIP part — you don’t hear the sorts of pathologies you get on cell phones from dropped packets, ie brief gaps, or squeals, but rather a high level of background noise or echoes or flanging, things that digital packets wouldn’t produce.

So the real issue is why one has to jump through all these phone card hoops to use VoIP, and as we all know, that’s a political, not a technical issue.

20

Dutch 09.24.04 at 8:48 am

One of the reasons of the failure of VOIP in the consumer market is the mobile phone. People are replacing their landlines with cellphones, not with VOIP phones.

Over where I live 89% of kids between 10 and 14 have their own cellphone. For the 14 year aged group that number is 98%! (That is no typo. Dutch only report.) I don’t know the figure for a (shared) land line but I would think that that is below 10% for that age group.

Why would they ever need a normal or VOIP phone?

21

Ian Whitchurch 09.24.04 at 9:16 am

I want to re-emphasise what Maynard said.

I run a baby Australian oil company, and some of the contractors we’re hiring to analyse data for us live in and work the US.

When I need to talk to them, I spend $10 on a phonecard, punch in some numbers, and are able to talk to our contractors for many, many hours.

When it runs out, I buy a new one.

From my end, it’s a regular phone call. From their end, it a regular phone call.

But it’s VOIP in the middle.

And if my VOIP provider goes broke, I’m out the princely sum of $10.

Ian Whitchurch

22

Russkie 09.24.04 at 11:38 am

As Neil Katin notes, VoIP (and packet switching generally) puts the intelligence in the end points rather than the network (this is called the “End-to-end principle” and is acknowledged to be the reason that the internet triumphed over other network architectures).

So-called progressives should prefer VoIP because it commoditizes the services provided by the phone company. Your ISP supplies a “big dumb pipe” – and your terminal adapter supports increasingly sophisticated features at little ongoing cost.

The main reason that VoIP is not free currently is that it still requires a gateway to the traditional phone network. Once everyone is using it, the price for VoIP should be the same as the price for email ie. zero.

Randolph Fritz is wrong regarding bandwidth reservations. Packet delay is less and less of an issue, because there is a glut of unused transatlantic optical cable, and because ISPs these days have to provide enough bandwidth resources to keep their users very happy all of the time. Experience has shown repeatedly that adding bandwidth is cheaper than adding signaling and realtime management of the bandwidth that you already have.

23

Russkie 09.24.04 at 11:56 am

Jake McGuire wrote:

On the other hand, computers are REALLY FAST these days, so brute-force methods – like figuring out how to route each individual packet of a phone conversation instead of figuring it out once at the start of the call – become practical.

This has been true at least since 1997 or so, when routing hardware became just as fast as “label switching” ie. ATM and later MPLS.

But the continuing expansion of routing tables in the “default-free” zone is considered dangerous by some people – and might give some advantage to label-switching MPLS.

24

Tom Hudson 09.24.04 at 1:07 pm

russkie writes:

(quote)
Randolph Fritz is wrong regarding bandwidth reservations. Packet delay is less and less of an issue, because there is a glut of unused transatlantic optical cable, and because ISPs these days have to provide enough bandwidth resources to keep their users very happy all of the time. Experience has shown repeatedly that adding bandwidth is cheaper than adding signaling and realtime management of the bandwidth that you already have.
(/quote)

I think I have to disagree. It doesn’t matter how much bandwidth our ISPs buy if, when I want to talk to somebody across town, the packets have to travel up to MAE-East in Washington, DC and then back down. I’m not familiar with the post-deregulation structure of the US phone network, but certainly the old design wouldn’t have had the hit from propagation and queueing delay but would have just gone through a local exchange or between two local exchanges. Peering on the Internet is too scattershot to guarantee performance.

There’s also a weakness in VoIP that doesn’t get addressed in many presentations: the phone network is designed for several nines of reliability, and the backbone Internet comes close, but the edge of the Internet doesn’t. My employer’s techs manage to hose our network connection at least once a month. The (land-line) phone network is also designed to be independent of the power grid, and the Internet isn’t.

(Thanks to everybody for giving me ideas for questions for the midterm I’m writing on this stuff right now.)

25

Russkie 09.24.04 at 1:47 pm

Tom Hudson:

I don’t understand your point about peering. The Internet (particular in the US) is resilient enough that routing inefficiencies won’t materially affect the quality of a voice call – ie. an extra couple of hops or 10-ms of latency is something that you won’t notice.

I’m far from the US and use a US VoIP provider. Most of the latency is at the ingress to the transatlantic link, and the several hops in the US add latency that’s negligible. Total latency for me is < 200 ms, and quality is comparable to a regular international call. Do you prefer MPLS to IP if you are concerned about extra hops? Your point about 5 nines is obviously correct.

26

Randolph Fritz 09.25.04 at 11:17 am

The heavy use of VOIP in international calling is due to the extra-ordinary over-pricing of the international circuit-switched telecomm network. So there, because of price, is another VOIP market. It is only carrying a tiny part of the international voice traffic, though, and a great deal more bandwidth would have to be allocated to the services if they became more popular. That there is a glut of transoceanic fiber has no relevance to this unless that fiber is made available to carry the IP traffic.

Internet congestion tends to make itself felt at the local level, where the traffic demands are highest, rather than long-haul. My consistent experience with US broadband residential internet service is one of usually high speeds, with occasional annoying long pauses.

27

hortense 09.26.04 at 3:04 am

Should anyone want an erudite discussion of the subject, see http://www.telepocalypse.net/ regularly.

Comments on this entry are closed.