TRAFFIC TRUNK PARAMETERS FOR VOICE TRANSPORT OVER

MPLS

A. Estepa, R. Estepa and J. Vozmediano

University of Sevilla

C/ Camino de los descubrimientos s/n

Keywords:

Voice transport, MPLS Trafﬁc Engineering, VoMPLS, VoIP.

Abstract:

Access nodes in NGN are likely to transport voice trafﬁc using MPLS Trafﬁc Trunks. The trafﬁc parame-

ters describing a Trafﬁc Trunk are basic to calculate the network resources to be allocated along the nodes

belonging to its corresponding Label-Switched-Path (LSP).

This paper provides an analytical model to estimate the lower limit of the bandwidth that needs to be allocated

to a TT loaded with a heterogeneous set of voice connections. Our model considers the effect of the Silence

Insertion Descriptor (SID) frames that a number of VoIP codecs currently use. Additionally, two transport

schemes are considered: VoIP and VoMPLS. The results, experimentally validated, quantify the beneﬁts of

VoMPLS over VoIP.

This work was supported in part by the Spanish Secre-

tar

ıa de Estado de Universidades y Educaci

on under

the project number TIC2003-04784-C02-02

1 INTRODUCTION

Voice transport over New Generation Networks

(NGN) will likely make use of QoS-supporting

packet-switching networks. Multi-Protocol-Label

Switching (MPL, 2001)(MPLS) is a packet forward-

ing technique that facilitates the creation of Label-

Switched-Paths (LSPs) and allows the use of trafﬁc

engineering needed to support the provision of QoS

at an optimal cost.

The trafﬁc engineering (TE, 1999) inherent capa-

bility of MPLS allows to dynamically route a set of

forwarding equivalence classes over a so-called Traf-

ﬁc Trunk (TT) which follows the most adequate path

according to its trafﬁc characteristics, available re-

sources in the network and administrative criteria. For

the remainder of this paper, a TT will be used to trans-

port a set of voice streams which demand the same

QoS and follow the same LSP between two Label-

Edge Routers (LERs) as indicated in ﬁgure 1.

In order to provide a TT with trafﬁc engineering

capabilities, the source LER needs to be aware of a

number of its characteristics. Among them, we are

interested in the trafﬁc parameters (e.g. mean and

peak bit-rates), which can be calculated from the traf-

ﬁc characterization of each voice stream belonging to

the TT. These trafﬁc parameters are required to calcu-

late the capacity to be reserved for the TT in each link

of the MPLS network and to develop a faithful map

of the overall capacity remaining free in the network.

Consequently the trafﬁc parameters are a basic input

to any constrained-routing algorithm.

MPLS

IP VoMPLS A2oMPLS

UDP

RTP

MPLS

IP VoMPLS A2oMPLS

UDP

RTP

To / From

Gateway Device

Link

Label Switching router

VoATM

...

others

VoIP

Voice Networks

PTSN

VoATM

...

others

VoIP

Voice Networks

MPLS Network

GW (LER)

Label Switched Path

(Label Edge Router)

Figure 1: Sample scenario.

The methods used to calculate the optimal capacity

to be reserved are usually based in complex analyti-

cal models (R. Gu

erin and Naghshineh, 1991) and are

out of the scope of this paper. However, the band-

width reservation should range from the sum of the

conversation’s mean bit-rates (stability condition) to

Estepa A., Estepa R. and Vozmediano J. (2006).

TRAFFIC TRUNK PARAMETERS FOR VOICE TRANSPORT OVER MPLS.

In Proceedings of the International Conference on Signal Processing and Multimedia Applications, pages 58-64

DOI: 10.5220/0001572800580064

 SciTePress

the sum of its peak bit-rates

. A common and simple

approach to calculate the actual bandwidth reserva-

tion is to use the sum of the conversation’s peak bit-

rates and take this upper limit as the allocation to be

requested for the TT. This guarantees that no packet

loss occurs at the cost of some over-provisioning of

network resources. However, this conservative peak

bit-rate approach could potentially cause the rejec-

tion of a new TT in the LER’s admission procedure

(CAC,) in spite of having enough capacity.

As the number of multiplexed sources in the TT in-

creases, the trafﬁc burstiness is smoothed and the ca-

pacity to be reserved should gradually change from

the peak bit-rate approach to the mean bit-rate ap-

proach, thus making a more effective use of the net-

work resources. However, the current calculus (B,

2002) of the mean-bit rate of the TT is inaccurate,

since it is based in the ON-OFF model which does

not consider the generation of Silence Insertion De-

scriptor (SID) frames that a number of voice codecs

generate during voice inactivity periods (Estepa et al.,

2003). These SID frames mark the end of talkspurts

and update the Comfort Noise Generation parameters

at the receiver.

Starting from the previous results in (Estepa et al.,

2003), we ﬁnd a more accurate analytical expressions

for the trafﬁc parameters of a TT (i.e. mean and peak

bit-rates) transporting a set of heterogeneous voice

sources when SID-capable codecs are used. We ap-

ply them to two possible voice transport schemes:

VoIP over MPLS and VoMPLS

. This would facil-

itate the use of the mean bit-rate value as a reference

for a effective resource allocation in trafﬁc engineer-

ing. In addition, the comparison between these differ-

ent transport schemes (i.e. VoIP and VoMPLS as ob-

served in ﬁgure 1) will let us to assess the bandwidth

savings of VoMPLS over VoIP. Our results could be

also applied to optimize the off-line analysis of packet

loss and delay by using the analytical models to pro-

vide a desired QoS level as a function of both the TT

mean bit-rate and the number of sources to be multi-

plexed.

The rest of the paper is structured as follows: sec-

tion 2 sets the basic models to transport voice over

an MPLS cloud and establishes the TT model used

throughout the paper. Section 3 calculates the maxi-

mum and minimum capacity allocation for a voice TT

in a VoIP over MPLS and VoMPLS scenario. Section

4 presents the main results and ﬁnally, section 5 con-

Within that range, the capacity selected represents a

balance between the maximum burst size and the probabil-

ity of out-of-proﬁle.

The case of A2oMPLS is not addressed in detail be-

cause the current implementation agreement does not spec-

ify the packetization scheme for the SID frames. How-

ever, the ﬁndings presented for VoMPLS are still valid for

A2oMPLS with some minimum changes

cludes the paper.

2 MODELS FOR VOICE

TRANSPORT IN MPLS

This section addresses two subjects: the characteri-

zation of a voice source trafﬁc a in a digital environ-

ment, and the means of transporting a set of those con-

versations belonging to a TT over an MPLS network.

Conversely to previous studies, we will not use the

ON-OFF model but the more general ON-SID model

presented in (Estepa et al., 2005). The main reason for

this is the inadequacy of the ON-OFF model to cap-

ture the effect of the SID frames in the conversation’s

mean bit-rate.

2.1 Single Voice Source Model: The

ON-SID Model

Low bit-rate codecs are commonly used in the trans-

port of voice over packet-switched networks. Typi-

cally, these type of codecs analyze the speech samples

generated during a period of time T and generate a in-

formation data-unit termed frame that can be used at

the receiver to faithfully restore the original sequence

of speech samples. Low bit-rate codecs are usually

equipped with a voice activity detection (VAD) fea-

ture which pursues bandwidth savings by avoiding the

generation of frames during voice inactivity periods.

Additionally, some audio codecs like G.729,

G.723.1 or AMR are also featured with an algorithm

which allows, at the beginning of each voice inactivity

period, to send SID frames. Reception of a SID frame

after a voice frame can be interpreted as an explicit

indication of the end of the talk-spurt. In addition,

SID frames may be also transmitted at any time dur-

ing the silence interval to update comfort noise gener-

ation parameters. This allows a faithful reproduction

of the background noise at the receiver’s side, increas-

ing the quality of the conversation at the cost of some

additional bandwith (Estepa et al., 2003).

Thus, the voice trafﬁc model to be used in the re-

mainder of this paper will not be limited to the tra-

ditional ON-OFF model, but the more general ON-

SID model. This model assumes that in the discrete

time space t

= i · T (where T is the codec’s frame

generation period), the codecs continuously generate

frames which can be either of type: ACT (compressed

voice), SID (background noise) or NoTX. The latter

corresponds to a zero-length frame used to model in-

stants when no frames (ACT nor SID) are being gen-

erated. ON and SID periods are exponentially distrib-

uted. During voice activity periods ACT frames are

generated every T seconds. During voice inactivity

TRAFFIC TRUNK PARAMETERS FOR VOICE TRANSPORT OVER MPLS

periods, SID frames are generated randomly accord-

ing to the codec’s speciﬁc algorithm and to changes in

the background-noise signal. Since SID frames gen-

eration is a random process, we can use a discrete

random variable, X, to indicate the inter-arrival time

(in number of periods T ) between SID frames as ex-

pressed in ﬁgure 2. Moreover, we assume that SID

frame generation is a renewal process.

S S S S

A A

...

Frames

SAAAA S S S SA

ON SID

T 2T

...

Packets

Figure 2: ON-SID frame generation model for VoIP and

f pp

=3.

Additionally, we also assume that during voice ac-

tivity periods, to compensate the excess of overhead

of layer protocols (H), ACT frames are usually sent

to the network in groups of N

fpp

consecutive frames

per packet. Note that this also causes a packetiza-

tion delay that limits the maximum acceptable value

of N

fpp

2.1.1 Mean Bit-rate of a Single Voice Source

According to the ON-SID model, a packetized voice

stream transmitted with VAD capable codecs which

transmit SID frames exhibits a mean bit rate of:

r = ρ · p + (1 − ρ) · r

SID

(1)

where ρ is the conversation mean activity rate, p is

the peak rate and r

SID

is the mean rate during voice

inactivity periods caused by the transmission of SID

frames. For those codecs which do not generate SID

frames, obviously r

SID

= 0.

The factors of equation 1 depend on the transport

scheme used (i.e. VoIP and VoMPLS,) and will be

addressed in next subsections.

2.2 Alternatives for Voice Transport

in MPLS

This paper addresses two possible ways of voice

transport over an MPLS TT, depending on whether

the tributary conversations come from VoIP or are di-

rectly taken from the payload of the VoIP packets; that

is, the transport of codec frames directly over MPLS

or VoMPLS.

2.2.1 VoMPLS

The implementation agreement deﬁned by the MPLS-

FrameRelay Alliance (VoM, 2001) describes how to

transport voice directly over MPLS. The method is

illustrated in ﬁgure 3 and can be summarized in the

following ideas:

• A number of voice calls may be transported over

an LSP. The multiplexing structure consists of a

mandatory Outer Label, zero or more Inner La-

bels, and one or more VoMPLS Primary Subframes

consisting of a 4-octet Header (HDR) and variable

length Primary Payload each, as shown in ﬁgure 3.

• Each Primary Subframe may be associated with a

different voice connection. A Primary Payload is

either a sequence of encoded ACT voice frame(s)

or a single SID frame.

• Within the header of a Primary Subframe, the

length ﬁeld is indicated in multiples of 4 octets.

Thus, up to three padding octets may be inserted

in each subframe depending on the codec’s frame

size and the number of codec’s frames carried in

the Primary Subframe.

• A Primary Payload contains the trafﬁc that is fun-

damental to the operation of a connection identiﬁed

by a Channel Identiﬁer (CID). It includes ACT and

SID frames. Primary Payloads are variable-length

subframes.

• Control Subframes may be sent to support the Pri-

mary Payload (e.g., dialled digits for a primary pay-

load of encoded voice) and other control functions

(like RTP-timestamps). Control and Primary Sub-

frames are not mixed together in the same multi-

plexing frame. Thus, Control Subframes will not

be considered in the present study since they be-

long to the signalling plane.

The header’s Channel ID (CID) allows up to 248

VoMPLS calls to be multiplexed within a single LSP

so the Inner Labels will not be considered in the rest

of the paper.

CID

PayLoad Type

Lenght

Counter

Inner Label . . .

C I D = Channel ID

H D R = Header

M = Mandatory

O = Optional

HDR

Outer Label

Subframe

4 octers

Primary PayLoad

(M) (O)

Primary Primary

Subframe

Primary

Subframe

Figure 3: VoMPLS trafﬁc trunk format.

The aforementioned implementation agreement

also establishes the maximum N

fpp

allowed value for

each codec (e.g. in VoMPLS there is a maximum

value of N

fpp

=6 for the G.729B codec, while for the

G.723.1 codec N

fpp

is forced to be 1).

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS

2.2.2 VoIP over MPLS

The protocols involved in the transport of IP packets

are the Real Time Protocol (RTP) and the UDP pro-

tocol, resulting in a RTP/UDP/IP header of 40 octets

per each IP packet.

Packets are generated every T · N

fpp

seconds dur-

ing voice activity periods. During voice inactiv-

ity periods, SID frames are packed according to the

RFC 3551 packetization scheme, where only those

SID frames consecutively generated may be carried

in the same packet, up to the maximum of N

fpp

Each VoIP stream is multiplexed in a Subframe of

the MPLS TT multi-frame structure. A new TT multi-

frame is sent whenever any conversation of the TT

needs to send a new IP packet. The multi-frame has

an Outer Label as indicated in next section.

2.3 Aggregation of Heterogeneous

Voice Sources in a Trafﬁc Trunk

For the aggregation of the voice sources in a single

TT we consider the multiplexing model illustrated in

Figure 4, where a set of K classes of m

homoge-

neous ON-SID sources feed a multiplexer (which can

be a LER) where a TT of real-time voice sources is

created. Each one of the m

voice streams belong-

ing to the same class i will have a common value of

fpp

and codec, and consequently, the same ON-SID

parameters; namely, the peak bit-rate (p

) and mean

bit-rate (r

m , r , p

1 1

m , r , p

K K KK

3 3

2 2

.m , r , p

MUX

L.E.R.

R , P , B

TT TT

Outer

Label

VoIP or VoMPLS

scheme frames

Figure 4: Voice multiplexing model.

According to ﬁgure 4, the trafﬁc characterization

should include the Outer Label of the TT. Therefore,

trafﬁc parameters deﬁning the trafﬁc proﬁle are:

• Trafﬁc Trunk’s Mean Bit-rate: this parameter is the

minimum service rate that guarantees stability in

the system and thus, is the minimum capacity that

should be allocated for the TT. It includes the sum

of all the conversations mean bit-rate for the class i

plus the mean bit-rate caused by the Outer Label of

the TT (R

T T

i=1

· r

+ R

(2)

• Trafﬁc Trunk’s Peak bit-rate: is the sum of all the

peak rates plus the peak bit-rate of the Outer La-

bel of the trunk supposed that all sources are ON

). This is the maximum capacity that should

be allocated to the TT to avoid packet loss.

T T

i=1

· p

+ P

(3)

• Trafﬁc Trunk burst-size: this parameter can be con-

sidered to be free. The reason for this is that the

allocated capacity (C) for the TT must be a value

greater than R

T T

(to be stable) and smaller than

T T

(to take advantage of statistical multiplexing).

For C=P

T T

, the buffer size (B) needs to be only

big enough to store one voice packet from each

conversation, while for C=R

T T

, B should be large

enough in order to queue all the instant trafﬁc in or-

der to bound the potential packet loss. An excellent

paper reviewing this tradeoff is (Procissi G, 2002).

Since the relation between B and the QoS depends

on the multiplexing analytical model used in the study

(i.e. either ﬂuid model or MMPP), our goal is to ﬁnd

the values of P

T T

and R

T T

for VoMPLS and VoIP

trunks. The next section is devoted to this task.

To account the Outer Label inﬂuence in the TT

mean bit rate, we make the following assumption: a

new MPLS frame is generated every T

min

=min{i =

1, ..k; T i} whenever there is any source generating a

new frame (i.e. voice frames or SID frames). For the

peak bit-rate calculation, we assume that a new trunk-

ing frame is generated every T

min

. Thus, the values

of P

and R

result as follow:

min

(4)

min

· G

T X

(5)

where G

T X

is the probability of having at least one

voice source generating a new frame at T

min

3 NEW VALUE OF THE TRAFFIC

PARAMETERS FOR MPLS

TRANSPORT

This section is devoted to ﬁnding out the analytical

expression for the trafﬁc parameters as indicated in

equations 3 and 2 of previous subsection. In our ap-

proach, we ﬁrst ﬁnd analytical expressions for p

and

for both transport schemes under study: VoIP over

MPLS, and VoMPLS.

TRAFFIC TRUNK PARAMETERS FOR VOICE TRANSPORT OVER MPLS

Table 1: Codec characteristics.

Codec Mode L

ACT

SID

T (ms) E[X] P

G.729 - 10 2 10 7.33 0

G.723.1 6.3 24 4 30 13.05 0.27

5.3 20 4 30

AMR 4.75 12 5 20 7.47 0

12.2 31 5 29 7.47 0

3.1 Mean and Peak Bit-rate for VoIP

The peak rate of a VoIP conversation depends upon

both the codec characteristics and the number of

frames per packet (N

fpp

). Thus, it is clearly given

by:

H + N

fpp

ACT

fpp

(6)

where H is the header size of the protocol layers in-

volved in the transport service (i.e. 40 octets), L

ACT

is the voice frame size and T is the frame generation

period of a given codec. Table 1 shows the character-

istics of some VoIP codecs.

Regarding the r

SID

member of equation 1, an ana-

lytical expression for the VoIP transport may be found

in (Estepa et al., 2005). The deduction was based

in the separation of the contribution of the header

and the SID frames to the mean bit-rate so r

SID

+ R

. The contribution of the SID frames can

be obtained by application of the Elementary Renewal

Theorem (ERT) which states that the SID frames ar-

rival long-term rate is the inverse of the expected

inter-arrival time (E[X] · T ).

SID

T · E[X]

(7)

where L

SID

is the size of a SID frame.

In VoIP, the contribution of the packet header gen-

erated during inactive periods follows the packet gen-

eration pattern imposed by the RFC 3551, where

one packet header is sent every non-consecutive SID

frame (X = x > 1). For consecutive SID frames,

one packet header is sent every N

fpp

frames, so both

cases must be considered. Since the mean time be-

tween SID frames is given by (E[X] · T ), the header

contribution (R

) can be expressed as:

= P

fpp

· T · E[X]

+(1−P

)

T · E[X]

(8)

where P

stands for the probability of having two

time-consecutive SID frames (i.e. P (X = 1)).

Thus, for the VoIP case we have an overall mean

bit-rate of:

= ρ · p

1 − ρ

T · E[X]



SID

+ H ·



1 +

(1 − N

fpp

)

fpp



(9)

3.2 Mean and Peak Bit-rate for

VoMPLS

When compared to the VoIP case introduced above,

VoMPLS transportation shows three main changes:

1. The header size (H) only accounts for one HDR

header, with a size of 4 octets instead of the VoIP

header of 40 octets.

2. The padding phenomenon may add extra octets to

the packets generated during voice activity periods

or SID periods.

3. The packetization scheme forces that one primary

subframe may carry only one SID frame. This im-

plies changes in R

when compared to the VoIP

case.

According to the ﬁrst and second items, an ex-

tra load of (N

fpp

ACT

) mod 4 octets needs to be

added in the ON periods. Thus, p

is:

H + N

fpp

ACT

+ (N

fpp

ACT

) mod 4

fpp

· T

(10)

On SID periods, an extra load of L

SID

mod 4

octets needs to be added to equation R

. Apply-

ing again the renewal theorem and taking into account

that only one SID frame can travel in the subframe,

we can redeﬁne:

SID

+ (L

SID

mod 4)

T · E[X]

(11)

and,

T · E[X]

(12)

So the mean bit-rate for VoMPLS will be:

= ρ · p

+ (1 − ρ) ·

H + L

SID

+ (L

SID

mod 4)

T · E[X]

(13)

where all the information units are measured in

octets.

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS

3.3 Trafﬁc Trunk Parameters

According to equations 2 and 3, the lower and upper

limits of the bandwidth reservation for the TT can be

readily calculated, since p

and r

have been deduced

in sections 3.1 and 3.2 for VoMPLS and for VoIP over

MPLS respectively.

However, the trafﬁc parameters of the TT should

also include the effect of the MPLS Outer Label in

the mean and peak bit-rates. To do this, we have to

compute the probability of having at least one voice

source generating a new frame at T

min

T X

). This

can be calculated from the probability that no source

from any class generates a packet.

In the VoIP over MPLS case, and according to

equation 8, it is given by:

T X

= 1 −

i=1

1 −

ρ · T

min

f pp

· T

(14)

−

min

(1 − ρ)

E[X

] · T

1 +

· (1 − N

f pp

)

f pp

In the VoMPLS case, we should consider that pack-

etization scheme forces that one primary subframe

may only carry one SID frame. Thus, G

T X

results

into:

T X

= 1 −

i=1



1 −

ρ · T

min

fpp

· T

−

(1 − ρ) · T

min

E[X

] · T



(15)

Experimental values show that, when more than 5

sources are multiplexed, G

T X

is greater than 0.95,

and for more than 10 sources, the probability in-

creases up to 0.99. Thus, we may assume G

T X

for a large number of multiplexed sources (i.e. more

than 10).

4 VALIDATION AND

NUMERICAL RESULTS

This section presents the results of a comparative

study between VoIP over MPLS and VoMPLS. This

allows to quantify the beneﬁts of VoMPLS, and vali-

dates the equations presented in the previous section.

4.1 Experiment Setup

Following the methodology found in (Estepa et al.,

2003), both edges of 14 conversations which took

place between males and females speakers, were

recorded from an ISDN line in a low-noise ofﬁce en-

vironment (i.e SNR > 20dB.) The 600 minutes of

PCM audio ﬁles obtained were encoded using the

G.729B codec. This codec, highly available in any

VoIP environment, holds the capability of generating

SID frames and is widely referenced in the literature,

so it will let us to compare our results with previ-

ous studies. The output of the codec was processed

to obtain the sequence of types of frames generated

(ACT, SID or NoTXN for none.) This information

was stored in a ﬁle -ftype ﬁles- for each conversation,

and all of them were processed to experimentally ﬁnd

the proper parameters to be used in the models (i.e.

activity rate ρ, E[X], P

The ftype ﬁles were also split into 120 pieces of

ﬁve-minutes-long conversations. This database of

ﬁve-minute pieces of speech conformed a pool from

which N were randomly chosen to feed a simulator.

Each simulation was repeated 40 times to provide ac-

curate measures of the TT mean bit-rate at the output

of the LER.

The mean bit-rate obtained in our simulations is

then compared to those provided by the analytical

expressions of R

T T

, for both VoIP over MPLS and

VoMPLS, respectively.

4.2 Numerical Results

A total of N=20 homogeneous voice sources were

multiplexed following the procedure explained above.

Figure 5 plots R

T T

as obtained from simulations and

from our analytical ON-SID model for both VoIP

and MPLS TTs. Additionally, it shows R

T T

when

SID

=0, as is traditionally assumed by the one-source

ON-OFF model.

In the VoIP over MPLS case the differences be-

tween our analytical ON-SID model prediction and

the simulation results, measured at the LER’s output,

range between 1% (N

fpp

=1) and 4% (N

fpp

=6). For

VoMPLS those differences vary from 0%(N

fpp

=1) to

5% (N

fpp

=6). This validates the analytical results in

both cases.

When using the traditional ON-OFF model, the

VoIP over MPLS case shows differences ranging

from 11% to 31%, at the same measuring point. In

the VoMPLS case, with the same frame-generation

model, the differences range between 11% and 19%.

This means that the SID-frames effect, known to be

non-negligible in VoIP, should also be taken into ac-

count in the VoMPLS case.

Note that, due to the padding phenomenon in VoM-

PLS, at N

fpp

=2 R

T T

is smaller than at N

fpp

=3. It

also demonstrates that N

fpp

=2 is an interesting work-

ing point for the G.729 codec, achieving less delay

with lower bandwidth consumption than N

fpp

=3.

Figure 6 reveals that using VoMPLS instead of

VoIP over MPLS yields bandwidth savings ranging

TRAFFIC TRUNK PARAMETERS FOR VOICE TRANSPORT OVER MPLS

Figure 5: Experimental and analytical values of R

T T

Figure 6: Bandwidth saving and ON-OFF error.

from 69% (N

fpp

=1) to 48% (N

fpp

=6).

5 CONCLUSIONS

Trafﬁc engineering needs accurate trafﬁc parameters

in order to calculate the optimal capacity allocation

for a Trafﬁc Trunk. We have provided analytical ex-

pressions for the mean and peak bit-rate of a Traf-

ﬁc Trunk loaded with a mix of heterogeneous voice

sources for both the VoIP and VoMPLS transport

models. Conversely to the ON-OFF based models,

the model used for voice sources captures the effect

of the SID frames generated by a number of modern

voice codecs. We show that the SID-frames effect has

to be considered in the VoMPLS case, too. This con-

veys an improvement in the accuracy of results, which

show a quantitative gain in the bandwidth necessary

to transport voice trunks when compared to VoIP.

The calculation of required bandwidth to be allo-

cated for a voice TT with QoS commitments is in

progress at the time of writing this paper. This sub-

ject as well as the A2oMPLS transport case are left

for further study.

REFERENCES

(1999). Requirements for trafﬁc engineering over mpls.

RFC 2702.

(2001). Multiprotocol label switching architecture. RFC

3031.

(2001). Voice over mpls- bearer transport implementation

agreement. MPLS Forum I.A.1.0.

B, G. (2002). Voice over internet protocol (voip). In Pro-

ceedings of the IEEE. IEEE Press.

Estepa, A., Estepa, R., and Vozmediano, J. (2003). Packeti-

zation and Silence Inﬂuence on VoIP Trafﬁc Proﬁles.

Lecture Notes in Computer Science, 2899(1):331–

339.

Estepa, A., Estepa, R., and Vozmediano, J. (2005). Accurate

prediction of voip trafﬁc mean bit rate. IEE Electronic

Letters, 8(10):644–647.

Procissi G, Garg A., G. M. S. M. (2002). Token bucket

characterization of long-range dependent trafﬁc. Com-

puter Communications. Ed. Elsevier, 25:1009–1017.

R. Gu

erin, H. A. and Naghshineh, M. (1991). Equivalent

capacity and its application to bandwith allocation in

high-speed networks. JSAC. IEEE, 9(7).

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS