Convergence – Voice Over IP Networks

• Subject Material
Š 11 Media Gateway Control Protocol (MGCP)
Š 11.1 Protocol Architecture
Š 11.1.1 Call Agent
Š 11.2 Signalling Gateway
Š 11.3 MGCP Protocol Overview
Š 11.4 Implementations
Š 11.5 Operation
Š 12 Quality of Service (QoS)
Š 12.1 Mean Opinion Score (MOS)
Š 12.2 Perceptual Analysis Measurement System (PAMS)
Š 12.3 Perceptive Speech Quality Measurement (PSQM
Š 12.4 Network Performance
Š 12.4.1 Network Performance Strategies
Š 12.4.1.1 Over Provisioning
Go To First Slide Slide 1
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Media Gateway Control Protocol (MGCP)
Š A Media Gateway (MG) will typically be a specialized
network element that performs a bridging function between
two dissimilar networks, such as the public switched
telephone network (PSTN), and a packet network.
Š The Avaya™ G700™ Media Gateway is such a device.
Š The MG will be required to provide the necessary
transcoding between the afore-mentioned dissimilar
networks.
Š For example, the G700 is capable of terminating Central
Office loop-start trunks, and converting the signalling and
analogue voice components for delivery onto a LAN.
Š It is also required to perform these operations in the
reverse direction.
Go To First Slide Slide 2
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP
Š The Media Gateway Control Protocol (MGCP) is an
architecture for controlling media gateways on Internet
Protocol (IP) networks and the PSTN.
Š In this architecture, telephone sets, such as VoIP sets, are
termed ‘endpoints’.
Š An endpoint is designed to implement and execute
commands under the direction of a device such as an H.323
gatekeeper.

Typical Exam Questions: Briefly describe the

function of a Media Gateway
What is an ‘endpoint’ in the MGCP architecture?

Go To First Slide Slide 3

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Protocol Architecture
Š Described in RFC 2805.
Š The current specific MGCP definition is RFC 3435.
Š The SDP is used by MGCP to specify and negotiate the
media streams to be transmitted for a call session.
Š MGCP then uses RTP for framing the media streams.
Š MGCP has a distributed architecture, and is composed of at
least three entities:
– a Call Agent (or Media Gateway Controller)
– at least one MG
– at least one Signalling gateway (SG) if connected to the PSTN

Typical Exam Questions: List 3 components of the

Go To First SlideMGCP architecture Slide 4
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Call Agent (CA)
Š The Media Gateway is controlled by the Call Agent.
Dialogue between the MG and Call agent is by way of MGCP.
The Call Agent uses MGCP to instruct the Media Gateway to:
– Report specific events to the Call Agent, such as dialed digits,
off hook, engaged.
– Connect endpoints.
– Play specific signals on endpoints.
– Report (audit) on the state of endpoints.

Typical Exam Questions: The Media Gateway is

controlled by the _________________
List 4 actions that the CA could instruct the MG to
Go To First Slideperform Slide 5
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Signalling Gateway
Š A Signalling Gateway is the network element that
transfers signal messages between Common Channel
Signalling (CCS) nodes where those nodes communicate
using dissimilar protocols and transports.
Š Quite often, the transport conversion will be IP to SS#7.
– Typically, these messages will contain information concerning
billing and the establishment (and teardown) of calls.
– Other messages may have information relating to location
requests, Global Title Translation, etc.
– The signalling gateway need not be a stand-alone element; it
can also be implemented as a component integrated with some
other network element.

Typical Exam Question: Define Signalling Gateway

Go To First Slide Slide 6
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Protocol Overview
Š Every issued MGCP command has a transaction ID and
receives a response.
Š There are 9 MGCP commands that start with a four-letter
‘verb’.
Š Responses begin with a three number response code.
Š Note that a Media Gateway could also send a DLCX if it
needs to delete a connection for its own management.

Go To First Slide Slide 7

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Protocol Overview MGCP Commands

Verb Used By Explanation Usage

AUEP Call Agent Audit Endpoint Query command – requests the state of an
MG

AUCX Call Agent Audit Connection Query command – requests the state of an
MG

CRCX Call Agent Create Connection Manages an RTP connection on an MG

DLCX Call Agent Delete Connection Manages an RTP connection on an MG

EPCF Call Agent Endpoint Configuration Modifies coding characteristics expected

by the "line-side" on the MG

MDCX Call Agent Modify Connection Manages an RTP connection on an MG

NTFY MG Notify Response for a previous RQNT from the

Call Agent that it has detected an event

RQNT Call Agent Request for Notification Requests notification of events on the MG
Requests an MG to apply signals

RSIP MG Restart In Progress Indicates to the Call Agent that the MG is

in the process of restarting
Go To First Slide Slide 8
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Protocol Overview
MGC --> MG CreateConnection: Creates a connection between two endpoints;
uses SDP to define the receive capabilities of the participating
endpoints.
MGC --> MG ModifyConnection: Modifies the properties of a connection; has
nearly the same parameters as the CreateConnection command.
MGC <--> MG DeleteConnection: Terminates a connection and collects statistics
on the execution of the connection.
MGC --> MG NotificationRequest: Requests the media gateway to send
notifications on the occurrence of specified events in an endpoint.
MGC <-- MG Notify: Informs the media gateway controller when observed events
occur.
MGC --> MG AuditEndpoint: Determines the status of an endpoint.
MGC --> MG AuditConnection: Retrieves the parameters related to a connection.
MGC <-- MG RestartInProgress: Signals that an endpoint or group of endpoints
is taken in or out of service.
Go To First Slide Slide 9
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Implementations
Š Two implementations of the Media Gateway Control Protocol
are in common use.
Š The names of both are abbreviations of the protocol group:
– MGCP is described in RFC 3435.
– Megaco is described in RFC 3525.
Š Although similar in architecture, MGCP and Megaco are
distinctly different protocols and are not interoperable.

Typical Exam Question: T/F MGCP and Megaco are

interoperable
Go To First Slide Slide 10
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Operation
Š When a gateway detects an off hook condition, it informs the
gateway controller.
Š The controller might respond with a command instructing the
gateway to put dial tone on the line and to monitor for DTMF
tones that indicate a dialled number.
Š After detecting the number, the gateway controller determines
how to route the call and, using an inter-gateway signalling
protocol such as SIP, H.323, or Q.BICC, contacts the terminating
controller.
Š The terminating controller could instruct the appropriate gateway
to ring the dialled line.
Š When the gateway detects the dialled line is off hook, both
gateways could be instructed by their respective gateway
controllers to establish two-way voice across the data network.
Go To First Slide Slide 11
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š MGCP Operation
Š In summary, these protocols have methods to:
– detect conditions on endpoints and notify the gateway controller
of their occurrence
– place signals (such as dial tone) on the line
– create media streams between endpoints on the gateway and
the data network, such as RTP streams.

Typical Exam Question: Summarize the operation of

the MGCP

Go To First Slide Slide 12

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Quality of Service (QoS)
Š I was once rumoured that the expectation from Access Providers
was a Central Office Switch that would have at most 40 minutes
of downtime in 40 years.
Š There is no current reference to that statement, but what is
generally accepted today is that core Telecom equipment should
meet ‘5 nines’ of reliability.
– That is to say, the equipment is in service 99.999% of the time.
Š The converse of this is that they might be down .001% of the
time.
Š We can calculate the annual downtime using this figure quite
easily: 365 days * 24 hours * 60 minutes * .00001 = 5.256
minutes; just over 5 and one quarter minutes per year.
Š That’s 5 times more than the 40 minute figure, but still within an
order of magnitude.
Š The bottom line is that the telecom industry has a high
expectation for the MTBF of its network equipment.

Go To First Slide Slide 13

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Quality of Service (QoS)
Š The big central office switches manage to meet this
requirement – they are large, fault tolerant computer
systems without a single point of failure.
Š Memory, cpu, disk, power are all backed up and redundant.
Š The weakest point is the software, but rigorous production
and testing methodologies produce systems that have
acceptable quality.

Typical Exam Questions: What is the weakest point

of a CO Switch?
Calculate the amount of downtime with ‘6 nines’
reliability.

Go To First Slide Slide 14

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Quality of Service (QoS)
Š Users of the traditional telecom network have grown accustomed to a
good level of service.
Š It is rare not to draw dial tone.
Š Long distance calls can be made with direct dial, and connect almost as
quickly as do local calls.
Š The difference between voice quality for a local or long distance call
should be indistinguishable – there is no technical reason for them not to
be.
Š Key measured voice quality metrics include:
– Clarity; a measure of how well and how clearly the reproduced speech duplicates
the original
• Effecting factors include silence suppression, CODECS, jitter, loss, and noise
– Echo
• The replay of transmitted ‘intelligence’ back to sender; increases with network
delay
– Delay
• End to end transmission time; manifested in drop-outs, static and packet loss

Typical Exam Questions: List 3 voice quality

Go To First Slide Slide 15
metrics.
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Quality of Service (QoS)
Š Mean Opinion Score (MOS)
– The mean opinion score (MOS) calculates a numerical value
that represents the perceived quality of a telephone
conversation.
– The MOS is expressed as a single number in the range 1 to 5,
where 1 is lowest perceived audio quality, and 5 is the highest
perceived audio quality measurement.
– In order to provide consistent results, the MOS tests for voice
are specified by ITU-T recommendation P.800
Š MOS was originally generated by providing human listeners
with a set of instructions detailing how to rate the audio
quality of the receiving end of a conversation.
Š The MOS is generated by averaging the results of test
sentences read aloud by both male and female speakers
over the communications medium being tested.
Go To First Slide Slide 16
Typical Exam Questions: Define ‘MOS’.
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Mean Opinion Score (MOS)
Š A listener is required to give each sentence a rating using
the following rating scheme:
MOS Quality Impairment
5 Excellent Imperceptible
4 Good Perceptible but not annoying
3 Fair Slightly annoying
2 Poor Annoying
1 Bad Very annoying

MOS Scores
Typical Exam Questions: T/F A MOS score of 1 represents
good quality
Go To First Slide Slide 17
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Mean Opinion Score (MOS)
Š The MOS is generated as the arithmetic mean (i.e. average)
score of the ratings of all of the participants.
Š We have previously seen that CODECS are available can
operate with less bandwidth than PCM/G7.11.
Š However, there is a trade-off between their operation and
voice quality as measured by MOS.

Go To First Slide Slide 18

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Mean Opinion Score (MOS)
Codec Data rate Mean opinion score

[kbit/s] (MOS)

G.711 (ISDN) 64 4.3

iLBC 15.2 4.14
AMR 12.2 4.14
G.729 8 3.92
G.723.1r63 6.3 3.9
GSM EFR 12.2 3.8
G.726 ADPCM 32 3.8
G.729a 8 3.7
G.723.1r53 5.3 3.65
GSM FR 12.2 3.5

Go To First Slide MOS Scores Per Codec Slide 19

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Mean Opinion Score (MOS)
Š MOS is indeed a subjective measurement, although ITU-T
P.800 makes a number of recommendations such as:
– Selection of raters
– Test environment
– Explanations to raters
– Analysis of Results
Š There are mechanized systems that can perform MOS
testing.
Š The result is that MOS yields similar but not identical results
for a given CODEC when done by humans.

Go To First Slide Slide 20

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Perceptual Analysis Measurement System
(PAMS)
Š PAMS uses a scoring method (from 1-5, 1 being the best) to
represent quality levels.
Š Each level on the points system from 1-5 represents a level
of quality by which the human ear can also detect.

Typical Exam Questions: Compare PAMS and MOS

Go To First Slide Slide 21

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Perceptual Analysis Measurement System
(PAMS)

Point Description
5 Very Poor Quality: Communication negligible, quality is so bad, the call is
made again
4 High Interference: Quality that makes the users change their normal
speech. Examples include: speaking louder, repeating
phrases
3 Noticeable Interference: Often irritating, gaps in conversation or jittering.
2 Noticeable Interference: That does not affect the communication quality of the
call.
1 Negligible interference: Excellent call quality resembling land line quality.

PAMS Measurement Scale

Go To First Slide Slide 22
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Perceptive Speech Quality Measurement (PSQM)
Š ITU-T P.861 provides an algorithm that attempts to faithfully represent
human judgment and perception with respect to voice quality.
Š This technique is referred to as the Perceptive Speech Quality Measurement
(PSQM).
Š PSQM generates a score that can be converted to MOS.
Š PSQM compares the output of a communications system to a known input
signal.
Š Included in the test are the combined perceptual effects of:
– Type of speaker (Male, female, child)
– Loudness of input signal
– Delay
– % of active / silent speech frames
– Clipping
– Environmental noise
Š The tests use a variety of speech samples that are ‘real’, or artificially
produced. ITU-T P.50 recommends speech samples for the artificial case.
Commercial test systems are available using PSQM, and are deployed by
developers of CODECS as well as by VoIP carriers.

Go To First Slide Slide 23

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š Telephone companies have used MOS for years as the
determinant for toll-grade calls.
Š If the MOS rating is above 4.0, then a call is said to be toll
grade.
Š The datacom industry has not used such a rating system
because data communication was never intended to be real
time.
Š In data communications, three measurements are used to
determine the quality of service.
– dropped packets
– jitter
Typical Exam Questions: List 3 Datacom
– latency QoS measurements.
Go To First Slide What is a ‘toll grade’ call? Slide 24
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š Real-time VoIP implementations have strict requirements in
order for the MOS to be at least toll grade:
Š Dropped Packets: depending on the CODEC, no more
than 10% average packet loss.
Š Some CODECs, such as G.723.1 require packet loss to be
less than 1%.
Š Since real time implementations require the use of UDP
rather the TCP, packet loss is a very real probability.
Š In addition, the loss must be spread out ‘nicely’ in order for
a CODEC to perform error recovery and correction and to
minimize the impact to voice quality.

Go To First Slide Slide 25

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š For example, assume the CODEC specification allows for 5
packets in 100 to be lost (5% packet loss).
Š However, if 5 packets in a row go missing, then it is unlikely
that the CODEC will be able to mask this loss, whereas if
about every 20th packet goes missing, there is a much
greater probability that the CODEC will be able to insert a
close guess as to what the missing packets would have
represented.
Š The most common causes of lost packets are router
discards, and late arriving packets.

Typical Exam Question: List the common

causes of lost network packets
Go To First Slide Slide 26
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š Latency: Also referred to as transit delay.
Š Transit delay is the total delay experienced by packets
containing encoded voice data as they travel through the
network; this is also referred to as mouth-to-ear delay.

Typical Exam Question: What is ‘mouth-to-

ear’ delay?

Go To First Slide Slide 27

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š A number of factors in the local and wide area network
contribute to transit delay:
– voice coding/compression Typical Exam Question:
– packet generation List 5 causes of transit
– channel contention (in a WLAN) delay
– network transport/buffering
– jitter removal.
Š If the one-way delay exceeds 150 ms, then the ‘cadence’,
or timing of the conversation will be adversely affected.
Š Distance, router buffering, and WLAN contention are all
contributing factors in end-to-end transit delay.

Go To First Slide Slide 28

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š Jitter: Jitter is a variation in delay.
Š In order to function, CODECs require a steady, constant
stream of data.
Š If there is jitter, then some data must first be buffered in a
‘jitter buffer’ before it is released to be processed by the
CODEC.
Š In this manner, the variable delay is ‘smoothed out’.
Š However, storing data in the jitter buffer also directly adds
to the latency.
Š If we configure the jitter buffer to store 100ms of voice, we
have already caused a 100ms transit delay.

Typical Exam Question: Define: Jitter

Go To First Slide Slide 29
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance

Š Echo Control: All telephone circuits introduce echo

(reflection of the speaker’s voice).
Š When the one-way transit delay exceeds 35 to 40 ms, the
echo becomes noticeable and annoying.
Š If the delay exceeds 40ms, equipment must be used to
remove the echo.
Š Virtually all packet voice networks will exceed 40 ms one-
way delay, so echo control will be one element that must be
incorporated into the design of a VoIP implementation.

Typical Exam Question: How much delay

will cause echo to be noticeable?
Go To First Slide Slide 30
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š Echo Control: All telephone circuits introduce echo
(reflection of the speaker’s voice).
Š When the one-way transit delay exceeds 35 to 40 ms, the
echo becomes noticeable and annoying.
Š If the delay exceeds 40ms, equipment must be used to
remove the echo.
Š Virtually all packet voice networks will exceed 40 ms one-
way delay, so echo control will be one element that must be
incorporated into the design of a VoIP implementation.

Typical Exam Question: How much delay

will cause echo to be noticeable?
Go To First Slide Slide 31
©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance
Š Other than meeting the ‘5-nines’ uptime expectation, the
major problem faced in implementing voice over data
networks is that most data networks were designed as ‘best
effort’.
Š This is not to say that data networks can’t be designed and
managed to provide toll-grade VoIP service. But what it
does imply is that in order to provide an end-to-end toll-
grade solution, then every element in the data path must
provide the bandwidth and transmission characteristics to
meet the delay, jitter, and packet loss requirements.

Go To First Slide Slide 32

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance

Requirement Characteristic of Internet

Voice Quality requires a minimum No guarantee of any minimum
bandwidth bandwidth
Voice traffic delays cannot exceed Amount of delay cannot be
300ms round trip guaranteed.
Low jitter Variable delay
Voice traffic needs high priority Traffic prioritization not generally
available
High Voice Quality Variable Quality

Go To First Slide Internet Characteristics Slide 33

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance Strategies
Š As a general rule, if traffic is to be treated specially, there
needs to be a mechanism to identify it.
Š In some cases, the application will have a configuration tool
that allows an administrator to mark or flag packets or
frames in some way as they are sent from the application.

Go To First Slide Slide 34

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance Strategies
Š Over Provisioning
Š Most layer-2 switches will run at wire speed between any
two ports.
Š Once a stream has been set up between two end devices on
the same switch, it is difficult to overload the switch to the
point where the relatively low bandwidth traffic (< 100Kb/s)
between the two ports can’t be serviced in real-time.
Š Even an older 10Mb/s switch shouldn’t be stressed in
supporting an internal VoIP stream.

Go To First Slide Slide 35

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance Strategies
A B

The problem will occur at an uplink

port, where all ‘off-switch’ traffic is
aggregated. This might be a link to
another layer-2 switch, or to the
gateway port on a layer-3 device such
as a router. It was mentioned that
Switch Y one of the driving factors for VoIP
Switch X

Uplink
was convergence, so that a mix of
real-time and non-real-time data
entering and exiting the switch is a
realistic scenario.
C D

Go To First Slide Slide 36

©2012 Prof Ian Miles
 Convergence – Voice Over IP Networks
Š Network Performance Strategies
Š Over Provisioning
Š One approach to resolving the above problem is to
overprovision the network – add extra capacity which is
seldom used.
Š For example, two separate LANs could be installed, one for
VoIP, the other for ‘regular’ use, and not allow traffic
between the two networks.
Š While more costly than programming the network devices to
meet quality of service requirements, it is a workable and
viable solution.

Go To First Slide Slide 37

©2012 Prof Ian Miles