Abstract
Being a BGP admin, I see an alarming tendency of telco operators to relate their telephony transport layers into the Internet. I personally consider this action inconsiderate and suppose there's a couple of issues to be discussed here.
| Pros | Contras |
|---|---|
|
|
Pros
An IP based structure makes more resources available for less money
The IP networks have no facilities to provide a guaranteed bandwidth to users. All clients simply share the available bandwidth on any segment of the network, which is normally equally distributed. Since most people don't really need the full 64 kilobit per second they could send, just giving them less bandwidth can help you fit more connections on the same amount of wires. Less wires make less money.
The Internet Protocol suite is less complicated than the PSTN UK0-, D-Channel and B-Channel protocools
The PSTN protocols themselves are relatively complicated. They need
to provide point-to-point connections, separate channels for data and
signalling and a guaranteed bandwidth (Which is constant per channel -
thank the Gods of the ETSI). However, the Common API for ISDN (CAPI)
provides for an easy-to-use interface to open a B channel you can send
data on, and close it afterwards. There are several applications for these
protocols, one of which is by the way the transmission of IP packets over
B channels.
On the other hand there is IP, which handles only a tiny fraction of the
service the PSTN protocol suite handles, which is transmission of data
over multiple endpoints. There is a certain amount of trust based
self-configuration via the Border Gateway protocol, but that requires
further interaction from administrators in most cases. The fact is,
IP is rather easy to implement, since it provides less features.
Contras
IP-based networks have proven themselves as unreliable
If you have a lot of permanent connections over multiple segments of the Internet, you will quickly realize the instability the Internet brings. Some hosts get flooded because their names appear on favorite news sites, some routes fall apart because someone accidentally misconfigured them, and from time to time large segments of the network simply fall apart, and it takes some time to fix the problems. It's debatable whether it's worth trading the well-known stability of the european PSTN for a minor decrease in implementation costs.
In cases of emergency, there is no guaranteed bandwidth
The PSTN has three facilities that provide for the ability to do phone
calls in case of emergency. The first one is the guaranteed bandwidth:
every B channel has exactly 64kbps, and there are no more B channels to be
allocated but the capacity of the line divided by 64 kilobit per second.
The other one is the precedence of alert calls: if someone calls the
police or the firefighters, less priorized calls will be cut off if no
bandwidth is available, so an emergency call is always granted. The third
facility is the own power supply: if you don't have any power at your
home, you can still do phone calls, because the telephones get their power
from the PSTN. This works even if the local power plants are shut down
or whatever.
The IP network, on the other hand, doesn't provide such functionality.
Connections share the bandwidth on any segment, and on each segment there
can be an unlimited account of connections. Thus, if you're trying to call
the police via VoIP, there is no guarantee whatsoever that none of the
traversed network segments will pass enough bandwidth to make you
understood. Remembering the fact that the PSTN operators have a duty
to allow emergency calls under all circumstances, the use of VoIP instead
of PSTN equipment may actually infringe on the law here.
In case of IP configuration problems, the PSTN may be required in order to fix the problem
Working for a carrier, I do see a problem with the switch. If the
IP network becomes unavailable, which happens about once per day in
average, what does one do? One calls up the operator of the unavailable
network and gets the problems ruled out with him. There is thus going to
be a significant problem if some important IP route becomes unavailable,
for example the line Zurich - London. The operators in Zurich are going
to be unable to call the operators in London, because the VoIP network
naturally became unavailable too, so there is a significant chicken-and-egg
problem trying to fix the network configuration.
Considering the fact that even most postal offices use IP to synchronize
their logistics, the only viable solution to this problem appears to
mount one's horse and take a ride to London.
The Internet is a model that relies on other countries and working connections to them
Most of the connections in the Internet rely on peer connectivity that
traverses frontiers. For example, if you want to connect from Switzerland
to Sweden, you'll likely traverse either Germany or Great Britain. In order
to access any of the root name servers, you will most likely require a
connection to London, so if that goes down, you're up a creek. This has
several effects: on the one hand, if your peer does some malconfiguration,
you cannot even tell him he did. As mentioned earlier, this happens daily.
On the other hand, this will significantly diminish the independence from
other states. Let's assume the british or US government gets corrupted,
as in someone would sneak into control by the way of manipulated elections
and start to take over the country, it's possible for any of those people
to just shut off internet access for the whole of Europe. This would also
lead to a situation where all PSTN connectivity will be gone, and any means
of communication will be impossible.
Luckily this situation is unimaginable with the current governments, err...
Customers cannot contact their service providers in case the network fails
It's a fact that the duty of the telcos is to maximize their shareholder value. However, this naturally involves customers, and caring about them. It's also a fact that customers like to be heard. However, if a problem occurrs in their internet connection, they usually call up your call center, and complain about it, so their urge to be heard is fulfilled. If however their phone connectivity goes down too, there's no way they can make themselves heard in the way they'd want to. Even though this actually reduces the costs on your call centers, it's probably not what your customers want. So it's an open question whether companies should care more about their customers or their shareholders.
IP-based networks need significantly more maintenance than the PSTN ever did
While the operational cost and production cost of PSTN devices may be slightly higher than that of IP networks, they require significantly more maintenance cost. In the PSTN, you have your points of presence, and you attach terminators to it that do the connection handling. If a network segment is overloaded, the traffic will regulate itself, because the number of connections per segment is limited, so a functionality is guaranteed. On IP networks there's a necessity to configure peering between autonomous systems, to ensure the routes are changed according to segment loads, to ensure that DDoS attacks have less of an impact, et cetera. You need people to do this work for you, and this takes money. Money that you're supposed to economize. On the other hand, the maintenance involved with the PSTN isn't very high, because most of the time it just configures itself.
Conclusions
On the basis of initial costs, VoIP solutions are quite good. However, the maintenance costs should quickly eat up what you economize, and in addition to that you end up with a chicken/egg problem the moment the connection is gone and you need help in order to fix it. VoIP is going to be cheap and great until it breaks and turns into a nightmare. And IP breaking down is a daily experience.