All rights reserved. December 2005.
The information in this document is subject to change without notice. The statements,
configurations, technical data, and recommendations in this document are believed to be
accurate and reliable, but are presented without express or implied warranty. Users must take full
responsibility for their applications of any products specified in this document. The information in
this document is proprietary to Nortel Networks Inc.
The software described in this document is furnished under a license agreement and may be
used only in accordance with the terms of that license.
Trademarks
Nortel, the Nortel logo, the Globemark, Unified Networks, PASSPORT and BayStack are
trademarks of Nortel Networks.
Adobe and Acrobat Reader are trademarks of Adobe Systems Incorporated.
All other Trademarks are the property of their respective owners.
Voice over Wireless LAN Solution Guide v1.0 December 2005
Abstract
This document is intended to define the Voice over Wireless LAN (VoWLAN) solution to assist
sales engineers in creating the best design to fit the customer’s environment while at the same
time eliminating common design errors. The products central to this document are the Nortel
WLAN Security Switch 2300 (models 2350, 2360, and 2380), Nortel WLAN Access Point 2330,
Nortel WLAN Handset (models 2210, 2211, and 2212), Nortel MCS 5100 Client, Nortel IP
Softphone 2050, and Mobile Voice Client 2050.
Voice over Wireless LAN Solution Guide v1.0 December 2005
1. Executive summary
Voice over Wireless LAN (VoWLAN) represents the coming together of two important and rapidly
growing technologies — WLAN and Internet Protocol (IP) Telephony. By seamlessly integrating
the IP Telephony system with WLAN infrastructure, VoWLAN provides users with high-quality
mobile voice and data communications throughout the workplace.
This document has two main purposes in defining the aspects of a VoWLA N produ ct solution.
Network designers need to know the engineering limits of various permutations of a network
design. However, this introduces a lot of complexity into the document, making it less useful to a
general audience. The primary purpose of this guide is to provide simple design
recommendations that resolve most issues and common scenarios, while not limiting the
applicability of the document. Therefore, the second purpose of this guide is to describe possible
deviations from the basic design that may be necessary in specific customer environm ents.
These scenarios present the most challenges and typically do not have a one-size-fits-all answer.
And so, in lieu of specific recommendations, these deviations are presented in a more openended manner that enables you to engineer customized solutions as necessary.
1.1 Challenges
Integrating voice applications on any data network poses some issues and challenges. WLANs
create a number of problems for voice above and beyond those inherent to most data networks.
This guide does not provide an exhaustive treatment of issues, but rather describes those that
are particular to supporting voice on 802.11 networks compared with typical data networks.
1.1.1 High overhead of 802.11
Unlike many other 802.n standards, 802.11 has a very high amount of overhead associ ated with
transmitting a packet. As a point of comparison, the difference in overhead for transmitting line
rate minimum frame sizes versus line rate maximum frame sizes on an 802.3 net work can be
significant, yet not nearly as significant as on an 802.11 network. For 802.11, the difference in
effective throughput varies dramatically with packet size due to the amount of overhead involved
in transmitting a frame. This means that the effective throughput of the medium is potentially
higher for data clients that use very large packet sizes than it is for voice clients that use smaller
payloads. As an example, using very conservative assumptions in terms of average frame size,
no rate scaling, and no contention or collisions, transmission overhead consumes as much as 67
percent of the total 802.11 medium capacity. Taking the same assumptions on an 802.3 network,
the overhead is by contrast about 8 percent.
1.1.2 Rate scaling and variable capacity
802.11b supports four transmission rates or data rates. Usually, as a client gets farther from an
Access Point (AP), both devices scale down to lower transmission rates in order to compensate
for a weaker signal. As a result, a transmission at the 5.5 megabits per second (Mbps) data rate
will take approximately twice as long as the same size packet transmitted at the 11 Mbps data
rate. That means less transmission time for other devices. Therefore, rate scaling compromises
the overall throughput of the medium. Rate scaling is necessary to extend the coverage of the AP
beyond a very tight region around the AP, but the effects should be taken into account when
determining medium capacity. For example, if the maximum call capacity for an AP is 12 when all
handsets are using the 11 Mbps physical (PHY) layer, then two handsets scaling down to 5.5
Mbps as they move away from the AP reduces the total call capacity of that AP to roughly 10.
This factor makes engineering the number of APs for the network difficult, because devices will
be roaming around and rate scaling up and down as necessary. Devices are moving, and the
engineering target of call capacity likewise becomes a moving target.
Voice over Wireless LAN Solution Guide v1.0 December 2005
1.1.3 Power adjustments and variable capacity
The WLAN market has matured to the point that most vendor product solutions have dynamic
mechanisms in place for adjusting channels, adjusting power, and filling coverage holes, all in
response to changes in the Radio Frequency (RF) environment. Although the robustness of the
mechanisms and features varies, all pose the same basic challenge to engineering voice
networks.
Dynamic adjustments work well for guaranteeing minimum coverage and connectivity of devices,
particularly data devices. Voice requires more deterministic engineering, though. Generally, the
number of calls per area (square foot) and calls per AP determines the number of APs required to
support the voice applications and devices. Yet power adjustments affect these parameters, for
better or for worse. If an AP increases power, it provides coverage for a larger area, meaning a
greater call demand per AP. Doubling the power of an AP may quadruple its coverage footprint,
which means up to four times as much call demand as originally engineered. As described in t he
previous section, that increased footprint will also have substantial portions of lower data rate
coverage. In addition, the added co-channel interference to other cells using the same channel
will degrade their call capacity. The net effect is that a network previously tuned for voice is now
less capable of meeting the demands of voice than it was before the dynamic power adjustment.
This is not to imply that auto-RF changes always have a negative impact on voice engineered
networks. Admission control techniques do help with the oversubscription problems related to
increasing cell sizes dynamically. Hole filling in case of AP failure also provides substantial value
to a voice solution. As a general statement, when VoWLAN is driving the engineering of the
network both in scale and capacity, sometimes auto-RF features create more challenges than
they resolve.
1.1.4 Quality of Service (QoS)
802.11 is a shared media technology. Only one device can use the media at a time. The AP
abides by this rule as well. Because collisions are impossible to detect by the transmitting device,
802.11 uses a statistical mechanism to reduce the possibility of collisions when two devices are
ready to transmit at the same time. When the medium becomes available, the mechanism
requires devices to wait a random amount of time before starting transmission. Because of this
simple mechanism, a non-voice device is equally likely to be allowed to transmit as a voice
device. If, for example, a data device does seize the medium, it could send a 1500 byte frame at
the lowest data rate (if it was far away from the AP), thus further delaying voice frames. In
addition, several data devices contending for the medium could each in turn send large frames
before the voice device gained access to the medium. Without a way to give preferential
transmission opportunities to voice devices as opposed to data devices, supporting voice
applications is a tremendous challenge on 802.11 WLANs.
SpectraLink Voice Priority (SVP) became the initial step towards QoS during the time when there
was no ratified standard for QoS, and evolved into a de facto standard for QoS, even though
SpectraLink handsets are the only terminals that support SVP. SVP does not solve all QoS
problems, but does serve as a model to illustrate the functions that a successful QoS mechanism
should implement.
The recently ratified 802.11e standard will ultimately resolve these QoS issues, but the delays in
the standard have created a number of additional implementation-specific challenges. Wi-Fi
Multimedia (WMM) is a step along the path to full 802.11e compliance for voice and multimedia,
not a solution, and because of this, QoS feature evolution will be marked by a progression
towards better and more solid standards-based QoS capabilities. WMM essentially refines the
existing statistical nature of 802.11 to give statistical preference to certain classes over other
classes. WMM is not a deterministic method of QoS. Because of this, it is full backward
compatible to legacy non-WMM devices, which function just like WMM best-effort class devices.
Voice over Wireless LAN Solution Guide v1.0 December 2005
QoS over the air techniques generally require complementary feature support by client and AP
alike, which means that some legacy devices or products that are slower to implement certain
features ultimately can impact the overall solution for voice with respect to QoS. The Hybrid
Coordination Function (HCF) is designed to smooth this transition by supporting a combination of
channel access methods, both new and legacy. However, it will be very challenging in the future
to attempt to support HCF Controlled Channel Access (HCCA)-based VoWLAN devices (future
802.11e), which are deterministic, alongside Enhanced Distributed Channel Access (EDCA)based (WMM) VoWLAN devices while providing sufficient quality to the latter group.
1.2 Scope of document
This solution guide does not encompass every WLAN product in the broader portfolio, nor does it
feature a number of LAN infrastructure components. The scope is defined as below.
1.2.1 Products Included
This solution guide addresses the following Nortel products:
WLAN Security Switch (WSS) 2300 series and WLAN Access Point 2330 (Release 4.0)
Ethernet Switch (ES) 460-24T-PWR and Ethernet Routing Switch (ERS) 5520
Ethernet Routing Switch (ERS) 8600 and Ethernet Routing Switch (ERS) 8300
WLAN Handsets 2210, 2211, and 2212
WLAN Telephony Manager 2245 and WLAN Application Gateway 2246
IP Softphone 2050 (PC)
Mobile Voice Client 2050 (PDA)
Multimedia Communication Server (MCS) 5100 and MCS Client
Communication Server 1000 (Release 4.0), Meridian 1 PBX (Release 4.0), and Business
Communications Manager (BCM) (Release 3.7)
1.2.2 Infrastructure components
This document covers in general terms topologies that include supported third-party AP products,
but does not address the configuration of those devices. In addition to the network equipment,
certain support devices, such as WLAN Management System (WMS) 2300, Enterprise Network
Management System (ENMS), Dynamic Host Configuration Protocol (DHCP) servers, and
Domain Name System (DNS) servers are addressed as well.
1.2.3 Configurations not included
This document does not currently address VoWLAN products in combination with WLAN Access
Point 2220/2221, WLAN Security Switch 2250, or Adaptive WLAN 2200 series. It does not
address networks that include Ethernet Switch (ES) 8100, Ethernet Routing Switch (ERS) 1424T,
or Ethernet Routing Switch (ERS) 1600 series.
2. Nortel VoWLAN solution
Due to the volume of material on the subject, the guide is organized in a topical manner. For
example, redundancy and security are treated separately, each imposing its own restrictions and
recommendations on the network. This section describes the basic architecture and some typical
design scenarios.
Voice over Wireless LAN Solution Guide v1.0 December 2005
2.1 Applications
Following is a brief description of the various voice applications.
2.1.1 WLAN Handset 2210/11/12 voice
The WLAN Handsets 2210, 2211, and 2212 work only in a Nortel Succession 3.0 (and later)
environment coordinated with a Communication Server (CS) 1000 or Meridian 1. These handsets
communicate with the Nortel call server through the Unified Network IP Stimulus (UNIStim)
protocol. The media path of the voice call goes from the handset directly to the destination device
(through the WLAN Telephony Manager 2245). In addition, the handset encapsulates all traffic in
the SpectraLink Voice Priority (SVP) protocol. The WLAN Telephony Manager (WTM) 2245
decapsulates the VoIP traffic from SVP and passes it onto the network—it does not translate
between UNIStim and SVP. Hence the WTM 2245 is in the path of all communication to and from
the handset. Likewise, signaling goes from handset to WTM 2245 to call server.
The WLAN Handset 2212 adds Virtual Private Network (VPN) capabilities to the handset
portfolio. It is a more durable version of the 2210 handset, and includes features such as backlit
display for night shifts and liquid resistance. The 2211 handset is the most durable and remains
the only handset in the portfolio that supports Push-to-Talk (PTT). The VPN features and PTT are
discussed later in this guide.
2.1.2 IP Softphone 2050
The IP Softphone 2050 is a voice application that runs on a regular PC or laptop. This has the
advantage of making the voice application itself decoupled from any particular hardware or radio.
Therefore, an IP Softphone 2050 can place calls over the higher-bandwidth 802.11a or 802.11g
network if the laptop has an 802.11a or 802.11g Network Interface Card (NIC). The Softphone
uses UNIStim for signaling to a Succession call server, but does not support SVP for QoS, and
consequently does not utilize the WTM 2245. The underlying device driver is responsible for
implementing QoS features such as WMM.
The IP Softphone 2050 is supported on the following operating systems:
Microsoft Windows 98
Microsoft Windows 98 SE
Microsoft Windows XP Professional
Microsoft Windows XP Home
Microsoft Windows 2000 Professional
Microsoft Windows 2000 Professional Service Pack 1
Microsoft Windows 2000 Professional Service Pack 2
2.1.3 Mobile Voice Client (MVC) 2050
The Mobile Voice Client (MVC) 2050 is also a UNIStim-based voice client that runs on a Pocket
PC Personal Digital Assistant (PDA). The MVC 2050 does not support SVP and consequently
does not utilize the WTM 2245. Signaling and voice bypass the WTM 2245, going directly to call
server and destination voice device respectively.
The MVC 2050 is supported on Microsoft Windows Mobile 2003 and Microsoft Windows Mobile
2003 SE on the following PDAs:
Multimedia Communication Server (MCS) 5100 is an application services delivery solution that
provides productivity, personalization, and collaborative applications that transform the way users
communicate. These SIP-enabled applications work together with an enterprise’s existing voice
and data infrastructure, evolving voice networks into true multimedia solutions, adapting to the
preferences of each user, so that communications become personalized, reducing repetitive timeconsuming tasks and streamlining business processes.
The MCS Client runs on a PC platform. The client uses Session Initiation Protocol (SIP) for
signaling to the MCS 5100 call server, but does not support SVP for QoS, and consequently does
not utilize the WTM 2245. The underlying device driver is responsible for implementing QoS
features such as WMM.
In the context of this document, the primary focus is the voice capabilities of the MCS Client.
2.1.5 WLAN Handset 2211 Push-to-Talk (PTT)
The WLAN Handset 2211 supports a walkie-talkie functionality called Push-to-Talk (PTT). The
PTT application utilizes IP multicast directly from handset to handset over the WLAN medium.
Each handset can be configured to be a member of one out of eight possible PTT “channels.”
The WLAN Handset 2211 joins the IP multicast stream through Internet Group Management
Protocol (IGMP) v1 reports. Each handset continues to send IGMP reports as long as PTT is
enabled, so the IP multicast state is maintained in the network even when there is no active use
of the PTT feature. The multicast address used to support PTT is 224.0.1.116, and each channel
uses a different User Datagram Protocol (UDP) port number from 5001 to 5008. For example,
handsets that are members of channel 2 would send the multicast to IP multicast group address
224.0.1.116 and UDP port 5002.
2.1.6 WLAN Handset 2210/11/12 text messaging
All WLAN handsets support text messaging applications through the WLAN Application Gateway
(WAG) 2246. The application server communicates to the WAG 2246 through a proprietary Open
Application Interface (OAI) messaging protocol. The WAG 2246 forwards the messages to the
WTM 2245, which encapsulates the message in SVP for delivery to the handset itself.
2.2 Network architecture
This section discusses some basic design principles that must be addresses before progressing
to more advanced topics. There are a number of deployment options for the solution as a whole,
but it is important to understand the basic design philosophy and basic design restrictions for the
solution.
2.2.1 Basic topologies
There are three basic network architectures and two AP connection types. Each switch in the
WLAN 2300 family usually enables one of the three architectures. The switches are not limited to
their respective roles, though it is generally a good practice to position the right switch in the right
A direct connection is defined as an AP with a physical connection to a WLAN Security Switch
2300 and which is configured as an extension to the physical port. This AP does not require an IP
address to function. A Distributed AP is an AP with a logical connection to a WSS 2300 over a
Layer 2 (L2) or Layer 3 (L3) network. It is controlled just as if it were directly connected through
the Control and Provisioning Protocol (CAPP). Do not confuse a Distributed AP (DAP) with
Distributed Campus architecture, as they are different terms. A Distributed Campus can
implement either direct connections or DAPs or both. Likewise, a Centralized Campus usu ally
implements DAPs but can also have direct connections if the WSS is a model other than the
WSS 2380.
DAP connections are always implemented as an L3 tunnel between AP and security switch, and
require the DAP to receive an IP address from a DHCP server. This means that whether the
physical topology is a routed network or an L2 network the DAP connection is the same. Put
differently, DAPs operate the same way whether there are routers or switches between the AP
and WSS 2300. However, when the DAP is separated from the WSS 2300 by a routed network,
then either DNS or DHCP option 43 (Vendor Specific Information) is required for the DAP to learn
the IP address of a WSS 2300.
2.2.1.1 Distributed Campus
In a campus environment, there are generally two choices for placement of an AP controller (that
is, WSS)—at the edge (in a wiring closet) or in a central data center. The Distributed Cam pus
represents the former choice. With security switches at the edge, APs can be directly connected
and powered by the WSS 2300. This basic architecture is shown in Figure 1.
The WSS 2300 models most often deployed in this architecture are the WSS 2360 and WSS
2361. They have six PoE ports and two network ports and support up to 12 total APs. A full
complement of APs would require six to be powered by another device (injector or Power over
Ethernet [PoE] switch) and configured as DAPs. Typically, the APs utilize a mix of direct
connections and DAPs
The advantage of a Distributed Campus architecture is that it leverages the integrated PoE ports,
which can be of value to a network that does not have much PoE capability in the wiring closets.
It also allows some of the Authentication, Authorization, and Accounting (AAA) features to be
leveraged for wired users through the Wired Authentication feature. Lastly, support of third-party
APs tends to suggest a Distributed Campus architecture as well, because the third-party AP must
be L2 connected to the WSS 2300.
Voice over Wireless LAN Solution Guide v1.0 December 2005
Figure 1: Distributed Campus architecture
2.2.1.2 Centralized Campus
A second architectural option is to centralize the security switches within a data center
environment. The model most suited for this role is the WSS 2380, which has four gigabit
interfaces, no PoE ports, and supports up to 120 active APs. Each AP is powered at the edge by
a PoE switch or a PoE injector, and has a DAP connection back to the central WSS 2300s.
Figure 2 depicts this architecture.
There are a number of advantages to centralization. Among them are the traditional benefits of
centralization, such as reduced numbers of manageable assets, lower operation costs, and
design simplicity. Within the context of the WLAN, there are additional benefits, such as simplified
security implementation (for example, fewer firewalls to deploy), simplified VLAN structure, and
simplified client management. With this architecture, you can also better leverage many other
Nortel core network features, such as Split MultiLink Trunking (SMLT) to a Nortel Switch Cluster
Core, and N-1 active-active resiliency capabilities.
Voice over Wireless LAN Solution Guide v1.0 December 2005
Figure 2: Centralized Campus architecture
2.2.1.3 Branch Office
The WLAN 2300 can also be deployed in a small branch office environment. Usually this type of
environment requires only a handful of APs, probably anywhere from one to six. The WSS 2350
is the model best suited for this environment, supporting up to three AP 2330s per WSS 2350,
including one PoE port for direct connection. A PoE injector or PoE switch would be required to
power the remaining two APs. The most cost-effective way to scale beyond three APs is to add
another WSS 2350. If you need more than six APs, you can also use the WSS 2360, if desired.
Figure 3 shows a sample Branch Office deployment.
There have been a few different industry approaches to supporting branch office environments,
such as deploying “fat APs” or “pseudo-thin APs.” The fat-AP approach packs all WLAN features
into the AP itself, but manageability is a significant problem. The pseudo-thin AP, which moves
many control functions back into the AP (making it semi-fat), attempts to provide the best of both
worlds in the branch but in practice tends to provide the biggest limitations of both worlds. All of
these approaches either complicate the support of the WLAN solution or impose severe
limitations on the branch APs. Putting a thin AP in the branch and controlling it over the WAN is
also not a viable solution, because most WANs lack the latency and throughput needs of the thinAP model. The best solution is to put a low scale security switch in the branch along with the APs
so that both controller and thin AP are collocated. This solution brings the full feature set to the
branch without adding extra complications.
Voice over Wireless LAN Solution Guide v1.0 December 2005
Figure 3: Branch Office architecture
2.2.1.4 Combining architectures
Up to now, architecture has been discussed in binary terms—this topology or that topology.
However, the WLAN 2300 solution is not restrictive in this way. The three architectures can be
combined in many different ways within the same network, and as will be shown later, VoWLAN
is generally not restricted by these architectural choices. However, it is a good practice from a
supportability standpoint to maintain a level of consistency in architectural choices. Figure 4
depicts all three architectures in one network, although this design example should not be
considered a reference architecture or a best practices diagram. In this network, both the
distributed and centralized WSSs are deployed as a single Mobility Domain, which is defined as a
collection of WSSs that work together to support a roaming user.
As a general rule, branch office WSSs are either in a separate Mobility Domain, if there are
multiple switches, or in a null Mobility Domain. They should not be part of the main campus
Mobility Domain, as users do not need seamless roaming between the sites. Also not desirable is
unnecessary loading of the WAN link due to the overhead associated with Mobility Domain
statistics collection and remote VLAN connectivity. When choosing whether to include the bran ch
office WSSs in the main campus Mobility Domain, base your decision on both geography (Do the
sites have seamless RF coverage between them?) and WAN speeds (Is there e nough bandwidth
to support users with remote VLAN assignments?). If you decide to backhaul a remote site’s
users to the campus Mobility Domain, you will need at least a T1 speed link, but you could require
higher speed depending on the numbers of users and application requirements. Within an
isolated branch office, if there is only one WSS, then no Mobility Domain needs to be defined. If
there is more than one WSS, then group them together in a unique Mobility Domain.
In general, most combined architectures are either Centralized Campus plus Branch Office or
Distributed Campus plus Branch Office. There is little motivation for combining Centralized
Campus and Distributed Campus in the same site, although there are some scenarios for mix i ng
architectures on individual sites. For example, a large campus may deploy WSS 2380s to support
the numbers of APs needed, a medium-size regional office may be big enough to deploy a
handful of WSS 2360s but not big enough to need a WSS 2380, and smaller branches might
employ WSS 2350s.
Voice over Wireless LAN Solution Guide v1.0 December 2005
Figure 4: Combined architecture
2.2.1.5 Third-party AP support
In some cases, integrating the WLAN 2300 series into an existing fat-AP deployment may be
required or desired. For instance, fat APs may have been deployed in a limited fashion and the
WLAN 2300 is a new network expansion. The WSS 2300 extends all its AAA features to the thirdparty AP, including the ability to consolidate all handsets into one voice VLAN/subnet, and QoS
classification capabilities. Figure 5 shows a sample third-party AP configuration.
Voice over Wireless LAN Solution Guide v1.0 December 2005
Figure 5: Network with third-party APs
The WSS 2300 supports many models of third-party APs with a few restrictions—see WSS 2300
Release Notes. All models of WSS 2300 have support for third-party APs, although
implementation specifics may vary. In many cases, Nortel also supports VoWLAN over third-party
APs. The APs must be L2 or directly attached to the WSS 2300. Note that this may require
extending VLANs in the Centralized Campus model using WSS 2380s. The APs must be certified
by SpectraLink as SVP compliant, and configured according to SpectraLink guidelines. You must
also meet WTM 2245 engineering requirements. Details about supported third-party APs, valid
network configurations, and other restrictions are beyond the scope of this document.
2.2.2 Network design constraints
While the previous sections discussed basic design types, the intent of the following topics is to
drill down to the next level of detail and explore various design and interoperability issues starting
with the physical layer (L1) moving up to L2 and L3 constraints.
2.2.2.1 Power over Ethernet (PoE) requirements
The AP 2330 draws an average of 6 Watts (W) of power when idle, increasing to 7 W average
under heavy load. The Ethernet Switch (ES) 460-24T-POE is capable of supplying up to a
maximum of 370 W of power (assuming ES 460-24T-POE is in AC+DC mode with Network
Energy Source [NES]) in some configurations. The ES 460-24T-POE with no secondary power
source is capable of supplying an aggregate of 200 W, which is enough to power a full
complement of 24 AP 2330s (assuming a draw of 7 W per AP).
Also, you must carefully consider redundant power options for the ES 460-24T-POE if PoE is
implemented. There are two external power supply options for the ES 460-24T-POE. The first is
the Ethernet Switch Power Supply Unit (PSU) 10. In the event of a primary AC power source loss,
Voice over Wireless LAN Solution Guide v1.0 December 2005
the ES PSU 10 can provide up to 75 W of power to PoE devices through the ES 460-24T-POE.
This means that when running on battery or redundant DC power, only about 10 or 11 APs can
be powered. If the deployment of the network calls for more than 10 APs on any ES 460-24TPOE switch, Nortel highly recommends the NES as the external power supply option instead of
the ES PSU 10. The NES can provide a full 200 W to end devices from a battery or secondary
power source in the event that the ES 460-24T-POE loses its AC power source. Therefore, the
NES offers full redundancy to 24 powered AP 2330s even in the event of loss of power.
2.2.2.2 Physical layer interconnections
The WSS 2380 has four dual PHY gigabit interfaces. You can use either the integrated
1000Base-T (copper) port or the Gigabit Interface Connector (GBIC) slot for fiber connectivity.
The only GBIC currently supported is a 1000Base-SX multimode GBIC with SC interface. You
can group all four ports into a MultiLink Trunk (MLT); the preferable mode of connection to an
Ethernet Routing Switch (ERS) 8600 core is to leverage SMLT on the ERS 8600 by connecting
two WSS 2380 ports to each ERS 8600. See Figure 6. All ports are active in this configuration, so
this represents high-throughput, highly resilient mode of network connectivity.
Note that while this document refers to MLT on the WSS 2300, the underlying link aggregation
technology is 802.3ad.
Figure 6: WSS 2380 using ERS 8600 SMLT
If the WSS 2380 is connected to an ES 470-24/48T, Business Policy Switch (BPS), or ES 460
equipped with a BPS-2000-2GE Media Dependent Adapter (MDA), then you must disable auto
negotiation on the WSS 2380. If you do not disable auto negotiation, the link will not be
established. When connecting to other ES products, including the BPS or ES 460 equipped with a
BS450-1SX/SR MDA, you must enable auto negotiation. The ERS 8600 does not have any auto
negotiation issues on gigabit fiber ports, so you must enable auto negotiation on the WSS 2380
when it is connected to an ERS 8600.
The WSS 2360 has six PoE ports and two regular network ports, all of which are 10/100 physical
ports. Because of this combination, the most recommended network connectivity option for the
WSS 2360 is to aggregate the two network ports into an MLT and connect to SMLT on an ERS
8600. The other six ports remain reserved for direct device connectivity requiring PoE, most likely
2330 APs.
A WSS 2350 is usually connected through its one network port to an L2 switch in the branch. One
AP can be directly connected to the other PoE port or indirectly connected through the L2 switch
as a DAP. Another variation that adds a little more resiliency is to aggregate both ports as an
Voice over Wireless LAN Solution Guide v1.0 December 2005
MLT to the L2 switch. In this configuration, all APs in the branch must be DAPs powered by the
L2 switch or a separate PoE injector.
In most cases, you should disable spanning tree on the switch port to which a DAP is connected.
If spanning tree is not disabled, there is a possibility that the AP will never connect to a WSS
2300 because of timing differences between spanning tree and the AP switch detection timers. In
the case of the ERS 8300, spanning tree does not put the port in forwarding mode before the AP
reboots to reinitiate a new connection attempt. The failed connection-reboot cycle repe ats
indefinitely until spanning tree is turned off.
2.2.2.3 RF design constraints
The WLAN Handset 2210/ 11/12 and most PDAs that support the MVC 2050 are currently
802.11b-only devices. Most APs shipping today are 802.11g capable, and the IP Softphone 2050
and MCS Client run on PCs that have 802.11b, 802.11g, and 802.11a interfaces readily available.
This creates some interesting dynamics and interesting choices for network deployments. The
following points lay the groundwork for a discussion of these choices.
1. Separation of devices by multiple Service Set Identifiers (SSID) on the same radio does
not create multiple shared mediums—the devices still transmit and receive using
common radio resources on a common channel.
2. Current QoS mechanisms in the industry are most effective at protecting and
prioritizing traffic on the downstream, that is, from AP to Mobile Unit (MU). Wi-Fi
Multimedia (WMM) improves upstream prioritization by giving a statistical edge to
different classes of devices so they are more likely to transmit ahead of lower class
devices. Still other devices may cheat on the contention window to gain a statistical
advantage, though there are drawbacks to this method. The bottom line is that there is
no real arbitration or coordination between multiple devices that need to transmit
packets upstream.
3. The 802.11g devices in a mixed 802.11b/g network are statistically favored by a 2:1
ratio over 802.11b devices. This means that if there is one 802.11g device and one
802.11b device and both are trying to saturate the medium with a data transfer, for
example, the 802.11g device will transmit, on average, two frames for every one frame
from the 802.11b device. If there are two 802.11g devices for every one 802.11b
device, then on average four 802.11g transmissions will occur before one 802.11b
transmission.
4. The 802.11g transmissions take less time than 802.11b transmissions due to the higher
data rates. So even though 802.11g devices transmit more often, they spend less time
transmitting packets, which mitigates the effect of 802.11g devices being favored
statistically from the perspective of the 802.11b clients. Having too many 802.11g
devices relative to 802.11b devices upsets this balance though, so beware.
Should you enable 802.11g or maintain an 802.11b-only network? There isn’t an easy answer. If
there is a significant amount of upstream traffic from data devices, then the question becomes
unimportant. The best course of action in that case is to keep data devices off the 802.11b/g
network entirely. Large numbers of 802.11g devices can also cause problems with 802.11b
handsets on the medium. However, if instead you make those 802.11g devices actually use
802.11b for communication, the situation will likely become worse. Disabling 802.11g support and
maintaining a dual mode 802.11a/b network can make 802.11a more attractive for dual mode
data clients and reduce the amount of data devices using the 2.4 GHz spectrum. Enabling
802.11g support may increase the number of data devices sharing the 2.4 GHz channels, which
is detrimental to voice devices. As a general policy, when you have large amounts of data, use
802.11a for data and 802.11b for voice, but leave 802.11g disabled.
On the other hand, if you only have a handful of 802.11b/g capable (non-802.11a capable) data
devices and the WLAN is to be used primarily for voice, then enabling 802.11g support is
Voice over Wireless LAN Solution Guide v1.0 December 2005
beneficial to overall voice quality and media scalability. These are some of the dynamics to
consider when making this choice. Ultimately you want to carefully control the number of data
devices sharing radio resources with voice devices, and you should gear your choices towards
this end.
For example, suppose that you have a large amount of Centrino laptops in the campus. If you
enable 802.11g mode, it becomes very likely that a large proportion of those laptops will prefer
802.11g (2.4 GHz) for connectivity, making it much more difficult to provide good quality voice for
handsets. If you disable 802.11g, those laptops will likely prefer 802.11a (5 GHz) because it
offers much higher throughput compared with 802.11b, and voice quality will benefit.
Another important constraint relates to maximum cell size for WLAN Handset 2210/11/12 voic e
support. Ideally, an AP has a circular coverage pattern, but in reality obstacles, antenna patterns,
and orientation occlude perfect RF patterns. So instead of listing cell size in terms of distance,
Received Signal Strength Indicator (RSSI) is used instead. The outer boundary for good voic e
coverage is anything better than –70 Decibels (dB). Therefore, strive to ensure that all areas of
the building are within the –70 dB range of some AP.
So far, the assumption has been that there is no interference on the 802.11b channels. But when
you deploy more than three APs, the APs themselves are a very important source of interference.
This is known as co-channel interference. Hence, it is also important to consider how channel
reuse (the term for channel plans in which more than one AP uses the same channel) impacts
network capacity. Generally speaking, you tile the channels to maximize the distance between
APs operating on the same channel. To scale capacity, you might add more APs in the same
geographic region while reducing the transmit power of each AP. But, the overall throughput
increase is not linear with the number of APs being added, meaning that total WLAN network
capacity is increased, but not proportional to the number of APs. This is an example of the law of
diminishing returns. The reason this happens is because each individual AP is losing throughput,
but the number of APs per square foot is increasing. Note that the biggest loss of per-AP
throughput occurs when going from non-channel-reuse to reusing channels. For more information
about this subject, see the
whitepaper available on the Nortel web site.
The goal of this RF engineering exercise is to achieve the required call density in terms of calls
per square foot. Getting the most calls per AP is not a useful objective of capacity planning. The
parameters that must be tuned in order to engineer a voice network for capacity are channel
reuse factor (that is, the number of channels in the channel plan), transmit power of each AP, and
radius of cell (that is, based on the physical distance between APs). Due to the complexity of this
topic and simulation data required, it is not possible to discuss tuning all three variables or even
two variables at a time. A simple case study of a light to medium office environment (mostly cube
space but some walls) is provided instead. The channel reuse factor for 802.11b networks is fixed
at three (three non-overlapping channels in the 2.4 GHz range), corresponding to channels 1, 6,
and 11, so this variable is set. We fix the transmit power to 50 mW, setting the next variable. Now
we compare the effects of cell size based on the other fixed parameters. When the deployed cells
have a radius of anywhere from 33 ft to 75 ft, the call capacity per square foot is essentially the
same. This means that packing cells in tighter than a 75 ft radius per AP is a waste of money.
This example shows that in a typical office environment with APs at half power, you should
deploy APs anywhere from 100 ft to 150 ft from each other. More walls mean you must have less
distance between APs, and lowering the power of the AP lessens the required distance betwee n
APs, both of which also serve to increase the net call density.
With respect to voice traffic from non-handset devices, calls may be placed over 802.11a or
802.11g networks. Because 802.11g has the same channel set as 802.11b, capacity planning
allows more calls per AP due to increased data rates, but the fundamental scaling limits and perAP radius limits remain mostly the same due to the limit of three non-overlapping channels.
Specifically, when only one channel is used capacity is much higher than 802.11b, but in an
enterprise deployment in which more than three APs are deployed and channels are reu sed,