8/12/2019 01628135

1/3

Bandwidth Provisioning and Capacity Planning Tools

J. Almhanaa, Z. Liu

a, R. McGorman

b, B. Lanteigne

a

aUniversity de Moncton, bLodex International Consulting{almhanaj, Lanteib, liuz}@umoncton.ca

Abstract

Bandwidth provisioning is an important issue for

Internet service providers (ISPs) and ensuring quality of

service (QoS) is a major concern. QoS is closely related

to the available bandwidth which itself is subject to

financial constraints. Unfortunately, up to now there havebeen no adequate tools available in the market so that

ISPs can do accurate bandwidth provisioning. In thispaper, we describe the software tools developed by

GRETI1 for that purpose. These Tools provide several

calculators for voice over IP and data traffic.

Keywords: Network capacity planning, Internet traffic,

Voice/IP, Traffic modeling, Network Tools

1. Introduction

For Internet service providers, bandwidth provisioning

and ensuring QoS are important issues. Especially whenfacing vigorous competition, ISPs must provide

satisfactory service levels, but cost pressures mean

provisioning must be economical. An ISP can face severalscenarios where more bandwidth is needed, including

growth in subscribers and increased user demands for

more Internet applications, such as VoIP (voice overinternet protocol) and other data traffic. Because of the

lack of appropriate planning tools, ISPs sometimes simplydouble network bandwidth and, therefore, their

investment costs. This practice, amplified by overly

optimistic bandwidth growth projections, led to a

catastrophic economical impact in the last 5 years. Some

cable companies now have more bandwidth than the

market can consume. In other scenarios, ISPs want toincrease profits from their existing communication

networks by adding users or other services but they maybe unsure about how to maintain QoS. QoS is related to

user satisfaction, and can have an economic impact when

dissatisfied users switch to another ISP for better service.

Some other scenarios also exist, for example, an ISP mayneed to establish pricing based on various levels of qualityof service.

1Groupe de Recherche en Technologies Avances dInternet

We believe that provisioning tools will help ISPs to

better manage their networks and optimize their resources.

Unfortunately, to the best of our knowledge, there havebeen no adequate tools available for bandwidth

provisioning and network capacity planning for ISPs whooffer so-called triple play services on the same medium:

VoIP, TV and high speed internet access. Some existing

VoIP tools [8] and video streaming calculators [10] offer

partial or incomplete solutions to bandwidth provisioning,but these tools are not capable of handling general internet

data traffic. Data traffic constitutes a major component ofnetwork traffic and must be taken into account. The

objective of this paper is to describe new software tools

we developed for bandwidth provisioning and network

capacity planning for integrated VoIP and data services.

These tools are part of the Data Traffic Analysis and

Tools Development project [12]. Our tools contain severalcalculators for bandwidth provisioning for a mediumcontaining VoIP and general data traffic. In this short

paper, we focus only on the description of the tools

without going into the implementation details, our main

purpose is to introduce our tools and show its importance

for ISP. The related theoretical background andimplementation details are documented in [2, 3, 4, 5].

The rest of the paper is structured as follow: Section 2

briefly explores some theoretical background. Section 3

describes our tools in more detail. Section 4 compares

our tools with various tools already available. Section 5concludes the paper.

2. Brief theoretical background of the tools

Traffic modeling is the first step towards bandwidthprovisioning and network resource optimization. As a

result, telecommunication traffic modeling has received

considerable attention during the past decade. For the

integrated service IP networks that we are considering, the

traffic consists of voice, high speed Internet data, video

streaming and background traffic. In practice, the totalbandwidths are divided into several parts and are

allocated to the various services. This static scheme iseasy to implement but wastes a lot of bandwidth. In this

paper, we focus on bandwidth provisioning and QoS for

voice and data services.

Proceedings of the 4th Annual Communication Networks and Services Research Conference (CNSR06)

0-7695-2578-4/06 $20.00 2006IEEE

8/12/2019 01628135

2/3

2.1. Available Models for Bandwidth

Provisioning

With an appropriate model, it is possible to perform

calculations relating bandwidth, number of users, and

QoS. For this reason, models are necessary for creatingprovisioning tools. Two well known models for voicetraffic are: Poisson often used in the US, and Erlang B,

often used in Canada. The main difference in their outputs

is that Poisson provisions more channels than Erlang B,

thus providing a margin of safety but also increasing the

possibility of having unused channels. There is also anextension to the Erlang B model, simply known as

Extended Erlang B. The Extended Erlang B takes into

consideration the fact that a caller who is blocked due to

lack of resources may retry immediately, increasing the

traffic load. Models such as Poisson and Erlang B can be

adapted for VoIP. This is accomplished by first finding

the number of channels needed, and then translating thisinto a bandwidth requirement based on the codec being

used and other factors. Another possibility is to use a

Gaussian approximation to the Poisson model to find the

number of channels.

Voice traffic was well studied and understood. The

challenge lies in modelling data traffic. A significantfinding was that data traffic has long memory, namely, the

autocorrelation function decays hyperbolically rather than

exponentially [10]. This means that traditional teletraffic

models based on Poisson processes are not suitable for

high speed telecommunication network traffic modeling.On the other hand, since long range dependent processes

have infinite memory, the model performance analysis

becomes very complex. For this reason a lot ofapproximate models have been proposed, including the

Markov modulated Poisson process models, e.g., [1] and

the M/G/infinity model [7].

2.2. Models used in the tools

For the purpose of bandwidth provisioning, the traffic

model should be very simple but flexible enough to

capture various traffic patterns. Based on the central limit

theory, when the number of subscribers is large enough,the aggregated voice traffic is approximated by a

Gaussian distribution. For VoIP, our Bandwidth Capacity

planning tools use the Gaussian model to perform thecalculations relating bandwidth, number of users and

blocking probability to determine the one unknown

quantity. On the other hand, when it comes to data trafficmodeling, our research has shown potential in the Gammadistribution, as seen in [5]. One of the data trafficcalculators we implemented uses the Gamma distribution

to perform its calculations. The other two data traffic

calculators use the dimensioning formula described in [3].

3. Description of our tools

Our tools include seven calculators, for which Figure 1

shows the main interface. Due to space limits, we do not

include the interface of each individual calculator.

Figure 1. The main interface

The first two calculators are dedicated to VoIP. The

first, called Voice Over IP Calculator, computes

bandwidth, number of users and blocking probability (i.e.QoS) for homogenous voice traffic using a single type of

CODEC. The second, called Multi-Codecs Voice Over

IP Calculator does similar computations but forheterogeneous voice traffic generated from several user

groups with different CODECs.For data traffic we have two options. The first is based

on dimensioning formula [3], which is implemented in

calculators 3 and 4. The second uses an empirical model

based on the Gamma distribution, and is implemented in

calculator 5.

Using the dimensioning formula, calculator 3, called

Dimensioning HSDT2 Calculator, computes one of thethree quantities: number of users, bandwidth, and QoS.

Given any two of these, it computes the third. Calculator

4, called Multi-Groups Dimensioning HSDT Calculator,

is similar to the previous one but with it we can defineseveral user groups with different requirements.

Calculator 5, called Empirical HSDT Calculator,

performs similar calculations to those of calculator 3 butusing the Gamma model [5].

2HSDT: High Speed Data Traffic

Proceedings of the 4th Annual Communication Networks and Services Research Conference (CNSR06)

0-7695-2578-4/06 $20.00 2006IEEE

8/12/2019 01628135

3/3

Calculators 6 and 7 are called Voice and Data

Calculator and Multi-Groups Voice and DataCalculator, respectively. With them, bandwidth

provisioning is done for traffic containing both data and

voice. Dimensioning formula [3] is only used to model thedata traffic.

4. Comparisons with other software

To the best of our knowledge, there are no similar toolsavailable in the market with which we can compare ourtools, especially for data traffic. However certain voice

calculators like [8] perform similar computations to our

first calculator. Our second calculator, related to voice

traffic, goes beyond simple bandwidth calculations by

adding multiple codec groups. This makes it more useful

in practice.

5. Conclusion

In this paper, the motivation and brief descriptions ofour Capacity Planning tools were given. We believe that

such tools are needed by ISPs and would help them

achieve better bandwidth provisioning in a market wherevigorous competition is growing continuously.

Our tools are currently at the prototype stage of

development. Further tuning in cooperation with ISPs and

more investigation are presently undergoing. In the future

we are planning to add video streaming to our tools.

Acknowledgments

This work was jointly funded by ACOA and Nortel

through the AIF to Dr. J. Almhana, and by NSERC to Dr.

V. Choulakian.

References

[1] Allan T. Andersen and Bo Friis Nielsen, A

Markovian Approach for Modeling Packet Traffic with

Long-Range Dependance,IEEE Journal, June 1998, pp

719-732.

[2] J. Almhana, Z. Liu, V. Choulakian and R. McGorman,

A Recursive Algorithm for Gamma Mixture Models,Proc. of IEEE ICC 2006, Istanbul, 2006.

[3] J. Almhana, Z. Liu, V. Choulakian and R. McGorman,

IP Network Traffic Modeling and Capacity Planning,Internal Report, 2005.

[4] R. McGorman, J. Almhana, V. Choulakian, Z. Liu, W.

Jedidi, Similarities between Voice and High SpeedInternet Traffic Provisioning,Proceedings of CNSR2004.

Fredericton, May 19-21, 2004.

[5] R. McGorman et. al. Empirical Bandwidth

Provisioning Models for High Speed Internet Traffic,Proceedings of CNSR2006, Moncton, May 24-25, 2006.

[6] K. Park and W. Willinger, Self-similar network traffic

and performance evaluation, John Wiley and Sons, Inc.

New York, 2001

[7] M. M. Krunz and A.M. Makowski, Modeling video

traffic using M/G/infinity input processes: A compromise

between markovian and LRD models,IEEE J. Select.

Areas Commun., 1998, pp. 733-748.

[8] Voice/IP Calculators http://www.voip-

calculator.com/calculator/ , 23/07/2004

[9] VoIP Providers List, VoIP Providers List VoIP

Calculator, http://www.voipproviderslist.com/voip-

calculator.

[10] VSS, Video Streaming Hosting Calculator,http://www.videostreamingservices.com/Hosting_prices_

calculator.htm, 2003.

[11] W. Leland, M. Taqqu, W. Willinger and D. Wilson,On the self-similar nature of Ethernet traffic,Proc.

ACM SIGCOMM'93, pp. 183-193, 1993.

[12] http://www.cnsr.info/research/componenth.php

Proceedings of the 4th Annual Communication Networks and Services Research Conference (CNSR06)

0-7695-2578-4/06 $20.00 2006IEEE