September 2002
Implementing VoIP Services In
The License-Free Wi-Fi
BY S.S. KAMAL
The title itself is a mouthful. But the unstoppable tide of
interest in wireless communication and the services it can carry is only
matched by the litany of confusing terms and expressions. So, to avoid
misleading the reader, the introduction to this article will be used to
define its focus. There are a lot of exciting things going on in voice
services and in wireless networks. This article doesn�t tackle them all.
Voice over IP (VoIP) is not always telephony; and all voice telephony is
not IP telephony. The terms �voice,� �IP,� and �telephony� can
be combined in many different ways to deliver services that are similar, yet
different. This article focuses on VoIP services, whether they are
telephony-based or not.
All wireless networks are not fixed (e.g., networks for cellular phones,
PDAs, and laptops roaming in an office or home.) This article focuses on
high-speed wireless networking between buildings and fixed structures,
covering neighborhoods, cities, and large service areas.
Finally, wireless frequencies are either licensed (i.e., their use is
exclusive to those who �own� the licenses for these frequencies in a
specific city or zone), or unlicensed (i.e., they are available for public
use). In both cases, these frequencies are regulated (government agencies
define how they are to be used, and certify equipment to verify that signal
and power rules are being complied with). Certification ensures everyone is
behaving, which is critical in the unlicensed frequency bands due to their
�public� nature. This article will focus on the unlicensed 2.4GHz
frequency bands.
The 2.4GHz frequency band has recently become very popular. Once derided
for being a �public� band (implying it was in some way polluted!), this
frequency band is currently being adopted by trendy �hot spot� operators
in airports, shopping centers, and coffee houses. In July this year, word
was leaked that large wireless operators were to conduct secret meetings to
discuss deployment of a national Wi-Fi network. It�s timely to examine
what VoIP demands of such initiatives... before it�s too late.
Building Blocks Of A VoIP Network
VoIP is increasingly viewed as a necessary element of advanced office
services. Its financial impact is both direct (in reduced telco costs), and
indirect (in operational productivity and new applications arising from the
convergence of an office�s separate voice and data networks onto a single
digital platform.
VoIP gateways or IP-PBX products play the role of �translator�
between the telephone world and the IP networking world. To the telephony
environment, they present standard telephony interfaces and features {analog
and digital interfaces, multi-tone dialing, IVR, etc.}. To the IP world,
voice and signaling information is transformed to IP packets that traverse
LANs and WANs accompanied by IP data, IP video, and the like.
Remote branch offices are typically �tied� to the main office through
the WAN. Although a remote office could also house a PBX, it can also make
use of another, smaller VoIP building block: the n-port VoIP adapter.
VoIP Demands On The Wireless WAN
Fixed wireless networks are a powerful option for the IP WAN. Not only
can they provide businesses a means of bypassing local telco facilities that
may not be able to deliver 10�50 Mbps connections, but also they free
businesses to establish their facilities anywhere, without worrying about
whether they can get connectivity. That�s true of all fixed wireless
infrastructure. It is especially appealing if unlicensed wireless networks
are used, since businesses can decide to set up their own private networks,
without resorting to any service provider.
Here we must pause to ask: �What effect will VoIP have on the wireless
WAN?� or �What effect will today�s wireless WANs have on VoIP?�
Those who have participated in an Internet telephone call have already
experienced the answer to both questions. The answer lies in the fact that
IP networks today are not yet designed for voice traffic (which is somewhat
of an ironic payback, since the 150-year-old telephone network still today
stymies the efficient transport of data). To no one�s surprise, the reason
behind this is how bandwidth in the network is managed.
How a device (any user device) accesses the network is a function of the
rules, protocols, and algorithms that allow the device access to the LAN
capacity, to which the device is directly connected, and then allows access
to the WAN medium that connects the device to the destination. For the WAN,
these rules reside in what is called the WAN media access layer (MAC). The
MAC rules typically straddle the second and third layers of the OSI protocol
stack. For the sake of this article, let�s say MAC-1 governs access to the
LAN. Access to the WAN is overseen by MAC-2.
The MAC layer in wireless networks varies widely due to many factors,
including what frequency band and physical layers you are adopting. In the
2.4GHz unlicensed frequency band, most outdoor wireless WAN suppliers have
adopted the IEEE 802.11(b).
The IEEE 802.11(b) MAC
The 802.11(b) standard, sometimes referred to as Wi-Fi, adopts a
protocol called CSMA/CA (Carrier Sense Multiple Access/ Collision
Avoidance). CSMA is a set of rules belonging to a whole family of protocols
called contention schemes. These are schemes by which parties access a
common, shared resource (like bandwidth) whenever they have the need to do
so. If more than one party decides that it needs to transmit at the same
time, collisions occur. In pure CSMA, these collisions are resolved by the
colliding parties stepping back and retransmitting again (and again if
collisions persist). In CSMA/CA the initial collision is avoided by having
the parties first �listen� to the channel they are all sharing. Only if
the channel appears free, will a listening party seize that channel and
begin transmitting. While this does not guarantee collisions are avoided, it
reduces them somewhat.
The problem with CSMA/CA is obvious from examining a wireless WAN
network.
If the REMOTE must listen before it transmits to the BASE, then which
REMOTE will likely get first notice of a free channel? The answer is: The
REMOTE that is physically nearest the BASE. The REMOTEs further away find
themselves unable to �seize� the channel, as their buffers overflow and
applications experience severe delays [imagine how telephone calls will
sound, originating from a REMOTE site].
As a wireless WAN grows in density, the REMOTE closest to the BASE may
not remain so. As the network expands in scope, more REMOTES will be further
and further away from the BASE, and at a severe disadvantage. This is
compounded by another weakness in 802.11(b) WANs, known as the �hidden
node� problem. If REMOTE sites cannot see each other, how do they �listen�
to a channel before deciding it is free and available?
The field results speak for themselves: CSMA-based networks start out
seemingly on solid ground. As REMOTE sites are added, the contention begins
to smother the channels in repetitive retransmissions; and REMOTEs further
from the BASE find themselves waiting endlessly, without access to the
network. Sites begin to transmit blindly. Frequently, this BASE site houses
central databases or gateways to the Internet that must be reached by all
the REMOTEs. Or the BASE is where the VoIP IPBX resides, and is the REMOTE
branch offices� only access to the public network.
Using CSMA in outdoor wireless WANs puts us in a limiting situation: we
either limit the network to lightly loaded traffic from sites that are all
sprinkled within somewhat equal distance of the BASE; or we overlay parallel
networks; meaning higher capital costs.
A question we should ask is: Why would IEEE 802.11(b) specify CSMA/CA, if
it suffers these shortcomings? The answer is that the 802.11(b)
specifications are optimized for indoor wireless LANs; not for outdoor
broadband WANs. Wireless LANs have the same topology as a WAN, but are used
indoors so that devices and laptops roaming inside a building need not be
tethered to CAT-5 cables. Wireless LANs use an Access Point [AP] attached to
the ceiling, instead of a BASE site; and the REMOTEs are plug-in or PC cards
that are inserted into laptops, and office equipment devices.
Inside buildings, distances between the AP and the roaming devices are
very short. All devices typically get simultaneous notice of the channel�s
availability. There are no far and near sites that easily confuse and
destabilize the CSMA/CA protocol.
Voice, like video and high-volume data services, cannot be left to �random
access protocols,� or to inconsistent channel qualities. Otherwise, the
connection can suddenly be seized by a third party in between talk spurts of
a conversation; or voice packets may incur long delays, as they wait for a
third party to release the channel for their services. Such intermittent
channel access results in�. intermittent� and� poor� voice�
quality.
Equipment suppliers who have adopted 802.11(b) for outdoor wireless WANs
try to �force fit� a solution by suggesting that (a) outdoor power
amplifiers can be added to the far REMOTE sites to lessen their handicap; or
(b) bandwidth shapers can be added to the LAN side to discipline the random
access to the channel. The first remedy can solve one problem but creates
another one: REMOTE sites with oversized amplifiers becoming an annoying
source of noise interference�and in a public frequency band no less. The
second remedy may help shape traffic overseen by MAC-1, but has very little
effect on the WAN access being governed by MAC-2.
Alternative MAC Implementations
Only recently have suppliers of wireless WAN equipment begun offering
non-contention MAC implementations. These have offered some relief for large
networks, or networks that carry heavy traffic (such as VoIP).
Polling Access: Also called round robin protocols. Here the BASE
station queries a REMOTE before allowing it to seize the channel. While this
applies discipline to the wireless network, it is not an intelligent agent,
i.e., it does not adapt to any traffic changes (such as idle REMOTEs with
nothing to send, or others that are swamped with traffic they must transmit
over the WAN).
Adaptive Access: This is a class of protocols that adapts to shifting
traffic patterns, and allow operators to set priorities to traffic on the
WAN (and sometimes on the LAN too). These protocols allows service operators
a great deal of flexibility in:
- Setting QoS levels to services.
- Guaranteeing various levels of �voice quality� to different
customers.
- Assigning fixed bandwidth to customers, mimicking leased wireline
services.
Market Forces Worth Considering
There are other design factors that affect VoIP being carried over fixed
unlicensed wireless networks, but the continued lack of adaptive bandwidth
management could stunt its wide deployment. In fact there are many experts
in this sector who predict that Wi-Fi will remain an Internet-access-only
medium, and confined to very small zones (like hot spots), as long as they
don�t become overpowered with user traffic.
There are reasons to be concerned about such a prediction. Firstly, it is
a step back from network convergence. Second, there is a strong move towards
blurring the lines between cellular services and Wi-Fi services (referred to
as portability) to allow users to move freely from office, to car, to home,
to lake house, etc. Such seamless integration cannot reconcile data-only and
voice-only segments. Our prediction is that such a force fit will result in
networks being cobbled together, a solution that can hardly be described as
seamless. c
Dr. S. S. Kamal is vice president, Solectek Corporation. Since the
1980s, Solectek has been one of the leaders in providing fixed wireless IP
products for true broadband networking between buildings, across towns, and
in cities around the world. For more information, please visit the company
online at www.solectek.com. To contact
the author directly, send e-mail to [email protected]
or call 858-450-1220 ext 3030.
[ Return
To The September 2002 Table Of Contents ]
|