P1: IML
Morris WL040/Bidgoli-Vol III-Ch-53 September 15, 2003 12:23 Char Count= 0
658 VOICE OVERINTERNETPROTOCOL(IP)depends upon several factors, including: (1) the nature of
the sound being transmitted, (2) the correlation in time
between the packets lost (i.e., are they bunched together
or widely dispersed in time), and (3) the randomness of the
losses. Because they do such an efficient job of squeezing
out redundancy, some CODECs that use data compression
algorithms are not able to recover from the loss of as few
as two packets in a row.
Jitter can also contribute to lower quality. With IP
traffic often taking multiple routes and/or mixed in with
bursty IP packet traffic, the interarrival times of the pack-
ets at the receiving end may be irregular and/or the
packets arrive out of order. Jitter can be overcome by
buffering the packets—i.e., temporarily storing them long
enough to reorder them before forwarding them to the de-
coder. However, buffering has the negative consequence
of adding more delay.
Delay in the transmission of voice packets can be mini-
mized in several ways. One is overbuilding the IP network,
i.e., ensuring that there is always excess capacity in the IP
network for all traffic, including during peak traffic peri-
ods. Alternatively, priority routing can be afforded to voice
traffic, at the expense of lower priority data traffic—which
is more tolerant of delays. Separate treatment of voice
from data can be done by virtual segregation or physical
segregation from lower priority traffic. The former can be
done at either the network or data link layer. For exam-
ple, segregation can be done at the data link layer by using
separate ATM channels for voice and data traffic and then
assigning ATM-based priority treatment to the ATM chan-
nels carrying voice traffic. Where voice and data are mixed
on the IP network, identifiers can be used to indicate to the
intermediate network components (such as routers) that
designated traffic (such as voice traffic) should be given
priority.
An example of this last method is RSVP, the Resource
Reservation Protocol. RSVP allows a VOIP application to
request end-to-end QoS guarantee from a network (Cisco,
VoIP Call Admission Control Using RSVP, 2003). If the
guarantee cannot be provided, the call will not be al-
lowed to go through. Where the guarantee cannot be se-
cured, the traffic might be redirected to an alternate net-
work or blocked (resulting in VOIP users receiving an
“equipment busy” signal). At the current time, priority
schemes such RSVP typically do not work over the public
Internet (with a large “I”). This is because, among other
reasons, the economic incentives are not there for inter-
mediate Internet providers to honor any type of priori-
tization routing scheme, given that their reimbursement
is the same for all traffic—regardless of its priority desig-
nation. Therefore the current market structure for public
Internet backbone routing prevents the realization of a
higher quality of service for VOIP traffic over the public
Internet.
To differentiate themselves, some Internet backbone
providers are introducing prioritization schemes such as
MPLS-based networks, which are ATM-like in their at-
tributes, but operate at a mixture of layer 2 and 3 proto-
cols in pure IP environments. As competition intensifies,
public networks are expected to become friendlier to real-
time services such as VOIP. Quality of service is discussed
in detail elsewhere.The Costs and Savings of Using VOIP
One source of VOIP’s cost savings over conventional tele-
phony is its ability to employ transmission more efficiently
due to both the extensive use of compression algorithms
and the statistical nature of its information transmission.
However, such efficiencies will tend to be most signifi-
cant to private networks and/or network providers whose
transmission networks are capacity-constrained.
A second source of VOIP’s cost savings is lower capital
cost per call using lower cost switching devices (i.e., In-
ternet routers and switches). Again, networks with sunk
investments in conventional technology with excess ca-
pacity would derive little benefit from such savings.
A third source of VOIP’s savings is lower costs of ad-
ministration, particularly in enterprise environments. A
good deal of administrative cost is incurred in enterprises
to accommodate the movement of telephone users within
the enterprise. Each time a user moves to another office
or enters or leaves the firm, the routing tables and direc-
tory of a conventional telephone system must be updated,
often manually. VOIP’s self-registration feature eliminates
these administrative costs.
A fourth source of VOIP’s savings comes from the
“economies of scope” that VOIP can achieve by its ability
to intermingle with other traffic on data networks, elim-
inating the need to segregate voice and data traffic, as is
often done with conventional telephony. These savings are
most easily exploited in the LAN and WAN enterprise envi-
ronment or by data-centric carriers who wish to combine
their voice and data traffic. Administrative savings also
come from eliminating the conventional regime of sepa-
rate administrative staffs for voice and data.
Finally, VOIP traffic on a data network looks like any
other data on that network. This allows some carriers and
enterprise users to avoid some of the economically dis-
torting taxes that local, federal, and foreign regulatory
regimes place on pure voice traffic, but not on data traffic.
It is important to remember that the realization of these
savings are application-specific and may not be realized in
every situation. See, e.g., Morris (1998) and “Cisco Seeks
Bigger Role in Phone Networks” (March 15, 2003).Security Issues for VOIP
Security is a concern with VOIP, particularly because of
the distributed nature of the call control, much of which is
handled between end devices (Cisco, SAFE: IP Telephony
Security in Depth, 2002). Some of the things that make
VOIP attractive, e.g., self-registration of end devices and
software-controlled PCs acting as end devices, are also
the sources of VOIP’s vulnerability. Security issues include
four categories: (1) eavesdropping, (2) toll fraud, (3) iden-
tity spoofing, and (4) IP spoofing.
Eavesdropping refers to an unauthorized party “lis-
tening” to the packets and, in turn, being able to listen
to the voice conversation. This problem exists with both
VOIP and conventional analog/digital voice communi-
cations. In both cases, the simplest method of prevent-
ing this problem is to encrypt the digital signal at its
source, the originating telephone. The problem encryp-
tion brings is overhead and computational load, which
can introduce delay. Also, encryption can create problems