Avaya and the Avaya Logo are trademarks of Avaya Inc. and may
be registered in certain jurisdictions. All trademarks identified by
® or ™ are registered trademarks and trademarks respectively of
Avaya Inc. All other trademarks are the property of their
respective owners.
NOTICE
While reasonable efforts were made to ensure
the information in this document was complete
and accurate at the time of printing, Avaya can
assume no responsibility for any errors.
Changes and corrections to the information
contained in this document may be incorporated
This document contains basic network requirements that are foundational for good voice
quality when using Avaya IP products and solutions over a data network. No document can
satisfy the detailed needs of every network, and therefore, this paper serves only as a
starting point. The document summary provides a short list of networking requirements,
allowances and recommendations. Use this page as a checklist to determine if the network
meets the minimum requirements for implementing Voice over Internet Protocol (VoIP) with
acceptable quality. The rest of the document contains basic networking and telephony
concepts for those who haven’t been exposed to a converged implementation. It also explains
why VoIP applications can yield poor results when data traffic on the same network doesn’t
seem to have problems.
Voice quality is always a subjective topic. Defining ―good‖ voice quality varies with business
needs, cultural differences, customer expectations, etc. The requirements below are based
on the ITU-T, EIA/TIA guidelines and extensive testing at Avaya Labs. Note that while
Avaya’s requirements will meet or exceed most customer quality expectations, the final
determination of acceptable voice quality lies with the customer’s definition of quality and the
design, implementation and monitoring of the end to end data network.
Quality is not measured by one discrete value where a number where 8 is good and 9 is bad.
There is a tradeoff between real-world limits and acceptable voice quality. Lower delay, jitter
and packet loss values can produce the best voice quality, but may also come with a cost to
upgrade the network infrastructure to get to the lower network values. Another real-world
limit is the inherent WAN delay over a trunk linking, for example, the U.S. West coast to
India. This link could easily add a fixed 150ms delay into the overall delay budget and is
beyond the control of an enterprise. Perfectly acceptable voice quality is attainable but will
not be ―toll‖ quality. Therefore, Avaya presents a tiered choice of values that make up the
requirements.
Voice quality is made up of both objective and subjective contributors. The objective
elements in assessing VoIP quality are delay, jitter and packet loss. These elements are
defined and influenced in the transport of IP both within and outside an enterprise. To ensure
good and consistent levels of voice quality, Avaya suggests the following network parameters.
Note that these suggestions hold true for LAN only and LAN/WAN connectivity. All
requirement values listed are measured between endpoints because this document assumes
that IP telephony has not yet been implemented. All values therefore reflect the network’s
performance without endpoint consideration. This is why there is, seemingly, a discrepancy
between the well-known ITU-T value for one-way delay and the values listed. The ITU-T
values are end-to-end values; from the mouth of the transmitter to the ear of the receiver.
The network requirements listed are meant for the network only – between endpoints - so
that your business data network can be assessed and modified, if need be, for successful
deployment of real time applications like voice and video. The requirement values are also
useful for ongoing network monitoring by IT staff. Upward trends in delay, jitter or packet
loss serve as a warning of potential voice quality problems.
Also, please note that ―Business Communication Quality‖ is defined as slightly less than toll
but far better than cell-phone quality. This is the tier where most businesses experience the
best trade-off between voice quality and network infrastructure costs.
o 80ms (milliseconds) delay or less can, but may not, yield toll quality.
o 80ms to 180ms delay can give business communication quality. This is far
better than cell-phone quality and in fact is very well suited for the majority of
businesses.
o Delays exceeding 180ms may still be quite acceptable depending on customer
expectations, analog trunks used, codec type, etc.
(See section 4.1 for more information)
Network jitter: Jitter is a measure of the variability of delay. Between endpoints:
o Toll quality suggests average jitter be less than ½ the packet payload. This
value has some latitude depending on the type of service the jitter buffer has in
relationship to other buffers, packet size used, etc.
(See section 4.2 for more information)
Network packet loss: The maximum loss of packets (or frames). Between
endpoints:
o 1% or less can yield toll quality depending on many factors.
o 3% or less should give Business communications quality. Again, this quality is
much better than cell-phone quality.
o More than 3% may be acceptable for voice but may interfere with signaling.
(See section 4.3 for more information)
Recommendations: Avaya highly recommends consideration of the following list of Best
Practices when implementing VoIP.
QoS/CoS: Quality of Service (QoS) for voice packets is obtained only after a Class of
Service (CoS) mechanism tags and Network elements treat voice packets as having
priority over data packets. Networks with periods of congestion can still provide
excellent voice quality when using a QoS/CoS policy. Switched networks may use
IEEE 802.1p/Q. Routed networks should use DSCP (DiffServ Code Points). Mixed
networks may use both as a best practice. Port priority can also be used to enhance
DiffServ and IEEE 802.1p/Q. Even networks with plentiful bandwidth should
implement CoS/QoS to protect voice communications from periods of unusual
congestion such as from a computer virus. See sections 3.1 - 3.4 for more information.
Switched Network:A fully switched LAN network is a network that allows full duplex
and full endpoint bandwidth for every endpoint that exists on that LAN. Although VoIP
systems can work in a shared (hubs or bussed) LAN, Avaya recommends the
consistently high results a switched network lends to VoIP.
Network Assessment: A Basic Network Readiness Assessment Offer from Avaya is
vital to a successful implementation of VoIP products and solutions. Contact the Avaya
representative or authorized dealer to review or certify your network. Section 7
―Network Assessment‖ explains the options available with this offer.
VLANs:Placing voice packets on a separate VLAN (subnet) from data packets is a
generally accepted practice to reduce both broadcast and data traffic from contending
for the same bandwidth as voice. Other benefits become available when using VLANs,
but there may be a substantial cost for initial administration and maintenance. Section
3.5 ―Using VLANs‖ further explains this concept.
Cautions: Avaya also recommends caution when using the following:
NAT: Be cautious when using NAT (Network Address Translation). Some
implementations using VoIP endpoints behind NAT fail because H.323 messages
contain multiple instances of the same IP address in a given message, and NAT can fail
to find and translate all of them. Avaya’s Communication Manager will work
seamlessly with any static NAT application even if that NAT is not H.323 aware. See
section 10.4, "Network Address translation" and Appendix C for more information on
using NAT.
Analog Dial-Up:Be careful using analog dial-up (bandwidth 56K) to connect two
locations. Upstream bandwidth is limited to a maximum of 56K, but in most cases is
less. This results in insufficient bandwidth to provide toll-quality voice. Some codecs
and network parameters provide connections that are acceptable, but consider each
connection individually.
VPN: Use Virtual Private Network (VPN) cautiously with VoIP applications. Older
systems can have large delays due to encryption, decryption and additional
encapsulation. Many hardware-based products encrypt at near wire speed and can be
used. Additionally, if the VPN routes over the Internet without SLAs in place, sufficient
quality for voice cannot be guaranteed unless delay, jitter and packet loss adhere to
the parameters listed above. See section 9.2 ―VPN (Virtual Private Network)‖ for more
information.
Voice over Internet Protocol (VoIP) is the convergence of traditional voice onto an IP data
network to provide better application integration by using a common protocol and to lower
costs by using ISPs and melding separate support staffs. Other real-time traffic, such as
uncompressed video and streaming audio, is also converging onto data networks.
VoIP is very complex because it involves components of both the data and voice worlds.
Historically, these worlds have used two different networks, two different support
organizations and two different philosophies. The voice network has always been separate
from the data network because of the protocols used and the characteristics of voice
applications are very different from those of data applications.
Traditionally, voice calls have had their own dedicated bandwidth throughout the circuit
switched network. This provided an environment where ―five nine‖ of reliability became the
standard. Interactive voice traffic is sensitive to delay and jitter but can tolerate some packet
loss, problems that were rarely an issue with circuit switching.
The data network, on the other hand, is packet switched. Data is less sensitive to delay and
jitter, but cannot tolerate loss.The data philosophy has centered on providing reliable data
transmission over unreliable media, almost regardless of delay. Bandwidth in the data world
is largely shared, so congestion and delay are often present and can cause problems for
multimedia applications such as voice.
The factors that affect the quality of data transmission are different from those affecting the
quality of voice transmission. For example, data is generally not affected by delay. Voice
transmissions, on the other hand, are degraded by relatively small amounts of delay and
cannot be retransmitted. Additionally, a tiny amount of packet (data) loss does not affect
voice quality at the receiver’s ear, but even a small loss of data can corrupt an entire file or
application. In some cases, introducing VoIP to a high performing data network can yield
very poor voice quality.
Therefore, implementing VoIP requires attention to many factors, including:
This document provides basic network guidelines to ensure good voice quality when
implementing VoIP. This document also examines some of the more important components
that affect VoIP and gives suggestions to help avoid problems during implementation.
3 Defining Quality
3.1 What is Quality?
Quality is a word that is used by almost all manufacturing and service providers. Quality,
however, is an ambiguous term representing superiority of that product or service. But
quality can mean different things to different people.
Consider Bill, a person who wants to buy a quality vehicle. Bill goes to a dealership and sees
a luxury sports sedan. It is a quality vehicle. The stitching on the leather seats is uniformly
0.2‖ apart on all seams. The finish consists of 10 color coats and 2 high-polymer sealant
coats. The fit between the doors and the body is consistently 0.167‖. Bill buys the luxury
sports sedan and is happy with the ―quality‖.
Now consider Trish, a person who also wants to buy a quality vehicle. Trish lives in rugged
mountain terrain, miles from anyone and must cross a boulder field just to get to work. Trish
is looking for a vehicle that has high ground clearance, a stiff suspension and 4-wheel drive to
get her to town and back – consistently without breaking down. Trish buys a Sport Utility
Vehicle and is happy with the ―quality‖. Trish doesn’t care about the stitching on the seats,
the gloss of the paint or the extreme exactness of the fit of the doors to the body. Trish
knows the paint will soon have chips, the body will get dents and the interior will stain.
Quality, in the above examples, consists of entirely different values. Therefore, what one
person values in quality may be almost irrelevant to another. Both Bill and Trish purchased
quality vehicles that have superior, but different features.
3.2 What is Voice Quality?
Defining voice quality is also difficult because the values of a small business can be greatly
different than a business that is larger or located in another culture or country. This is why
Avaya presents choices using a tiered system of network requirements. One number for
delay or jitter or packet loss cannot satisfy all customers in all businesses and in all cultures.
Ultimately, each business must decide if quality voice using VoIP requires the first tier of
values or other tiered values specified in this paper.
4 Prioritizing Voice Traffic
In order for a VoIP solution to function well, the network must be able to give voice packets
priority over ordinary data packets and sufficient bandwidth must always be available.
Avaya’s products for VoIP—Communication Manager™ Software include several standard
strategies to prioritize voice traffic. These strategies include using class of service (CoS),
prioritizing ports, prioritizing services, and using IEEE 802.1p/Q to set the priority bits. Avaya
products are designed to work with most other popular switches and routers using open
standards to provide end-to-end voice prioritization.
4.1 Understanding the difference between CoS and QoS
Class of Service (CoS) is a classification method only. CoS does NOT ensure a level of Quality
of Service (QoS), but is the method used by queuing mechanisms to limit delay and other
factors to improve QoS. Most CoS strategies assign a priority level, usually 0–7 or 0-63, to a
frame or packet respectively. Common CoS models include the IP TOS (Type Of Service)
byte, Differentiated Services Code Point (DiffServ or DSCP, defined in RFC 2474 and others)
and the IEEE 802.1p/Q.
Quality of Service (QoS) involves giving preferential treatment through queuing, bandwidth
reservation, or other methods based on attributes of the packet, such as CoS priority. A
service quality is then negotiated. Examples of QoS are CBWFQ (Class Based Weighted Fair
Queuing), RSVP (RESERVATION Protocol - RFC 2205), MPLS, (Multi Protocol Label Switching RFC 1117 and others).
CoS, or tagging, is totally ineffective in the absence of QoS because it can only mark data.
QoS relies on those tags or filters to give priority to data streams.
4.2 Using Ports
One prioritization scheme assigns priority based on the UDP (User Datagram Protocol) port
numbers used by the voice packets. This scheme allows the use of network equipment to
prioritize all packets from a port range. UDP is used to transport voice packets through the
LAN because, unlike TCP, it is not connection-based. Because of the human ear’s sensitivity
to delay, it is better to drop packets rather than retransmit voice in a real time environment
so a connectionless protocol is preferable to a connection-based protocol. By using
Communication Manager Software, users can define a UDP port range for voice priority.
Routers and layer 3 data switches can then use these ports to distinguish priority traffic. This
priority traffic can be voice packets (UDP), signaling packets (TCP) or both. This is an OSI
model layer-4 solution and works on data coming to and from the specified ports or a port
range.
4.3 Using DSCP (or TOS)
The DSCP prioritization scheme redefined the original Type of Service (TOS) byte in the IP
header by combining the first six bits into 64 possible combinations. This use of the TOS byte
is used by Communication Manager Software, IP Telephones, and other network elements
such as routers and switches in the LAN and WAN. A DSCP of 46 (101110) is suggested for
the expedited forwarding of voice packets. However, with Communication Manager, one can
set any DSCP value as desired to work with a company’s QoS scheme.
Note that older routers may require a DSCP setting of 40 (101000), which is backward
compatible to the original TOS byte definition of critical. But again, Avaya products and
software allows users to set any of the 64 possible DSCP values to work with your voice
quality policy. The TOS byte is an OSI model layer-3 solution and works on IP packets on the
LAN and possibly the WAN depending on the service provider.
4.4 Using IEEE 802.1 p/Q
Yet another prioritization scheme is the IEEE 802.1Q standard, which uses four bytes to
augment the layer-2 header. IEEE 802.1Q defines the open standard for VLAN tagging. Two
bytes house 12 bits used to tag each frame with a VLAN identification number. The
IEEE 802.1p standard uses three of the remaining bits in the 802.1Q header to assign one of
eight different classes of service. Again, with Communication Manager Software, users can
add the 802.1Q bytes and set the priority bits as desired. Avaya suggests you use a priority of