A PBX-BASED APPROACH FOR TELECOMMUNICATION

NUMBER PORTABILITY SERVICE

C. W. Cheng

Department of Computer Science and Information Engineering, National Chiao Tung University, Hsin-Chu, Taiwan

C. G. Chung

Department of Computer Science and Information Engineering, Chang Gung University, Kwei-Shan Tao-Yuan,Taiwan

Keywords: Number portability, PBX, cache policy.

Abstract: Number portability (NP) servic

e has become an important factor in enhancing the competitiveness of a

service provider. Although NP service is available now, NP database (NPDB) query and portable number

translation are all-call-based, and the time required to process a NP call is long. It is an overhead to the

operator and the user. How to reduce the impact of NP call process is an important issue. Considering an

organization often has dialled number locality, we suggest applying caches to PBX to keep the routing

information of portable numbers. In this paper, the interoperation of PBX and operator network is

introduced. We also demonstrate the workload of portable number translation is shared out to PBX, the

delay time of querying routing information is shorten, and the efficiency of NP call process is enhanced.

1 INTRODUCTION

The service of number portability (NP) that allows a

subscriber to keep a unique telephone number even

when changing service providers is an important

factor for fair competition among telecommunication

operators (Gans, King & Woodbridge 2001).

At present, approaches to support number

ortability in switching circuit networks (SCN) are

often implemented on the Intelligent Network (IN)

architecture (Lin & Rao, 2003). A number portability

database (NPDB) is required for an operator to

maintain the mapping between portable numbers and

destination addresses. Number translation is

all-call-based. Every NP call requires a NPDB query

for portable number translation. In addition, when a

number was ported out, the whole cluster of numbers

were taken as portable numbers. With the increase of

NP subscribers, NPDB grows cumbersomely, the

workload of NPDB is heavy, and the delay time for

NPDB query increases to prolong the call setup

process.

Many researches tended to enhance the call setup

ficiency by improving the performance of database

query. (Carpenter et al, 2000) pointed out that caches

can offload a substantial amount of traffic at a

database. To alleviate the workload of NPDB, a

cache keeping the destination address of portable

numbers can be applied to the system to reduce the

delay time of portable number translation. But

implying caches to the NP service still has problems.

Once a cache is implemented to the switching

center

of the operator, the numbers dialed from the

numerous subscribers are scattered. To increase the

cache hit ratio, the cache size must be large to

accommodate the amount of dialed numbers. This

could burden the effort of cache query and

management. If the cache is applied to terminals, the

altered routing information of portable numbers must

be updated from operators to all terminals; that may

consume considerable communication resources, but

the service provider makes no profit from it.

Caches profit when the variation of data is

occasional

or the accessed data has locality. From

the telecommunication network hierarchy, we found

that a lot of telecommunication traffic was generated

by organizations. Organizations usually have sets of

frequent contact targets, thus the locality of the

dialed numbers is distinguishable. When the

frequently dialed numbers are portable numbers, the

total time delay for waiting NP call setup is

137

W. Cheng C. and G. Chung C. (2005).

A PBX-BASED APPROACH FOR TELECOMMUNICATION NUMBER PORTABILITY SERVICE.

In Proceedings of the Second International Conference on e-Business and Telecommunication Networks, pages 137-142

DOI: 10.5220/0001417601370142

 SciTePress

appreciable. Moreover, every NP call setup requires

NPDB queries and consumes extra communication

resources, but service providers rarely make profit

from the process. Organization usually established

PBX (Private Branch Exchanges) for saving

telecommunication cost. We propose to apply caches

to PBX. By evaluating the costs and benefits of

applying caches to PBX, we show that PBX with

cache is a reasonable solution that can remarkably

alleviate the traffic and workload of databases and

improve the efficiency of NP services.

The rest of this paper is organized as follow.

Section 2 gives an overview of related work. Section

3 introduces caches to the telecommunication system

to enhance the efficiency of NP call process. Section

4 investigates the performance of caches in NP

service. Finally, section 5 draws the conclusion.

2 RELATED WORK

NP implementation schemes can be classified into

on-switch and off-switch solutions (Black 1998).

Methods of on-switch solutions implement routing

knowledge on switching centers of the service

network. When an originating network, which a

caller connects to, receives a call, it routes the call to

the donor network, which first assigned the

telephone number to the callee, by the prefix of the

dialed number. The donor network routes the call to

the destination network by the information of the

mapping of dialed numbers and the destination

addresses in the gateway switches. The operation

logic of a switching center alters whenever a number

is ported out or in. The frequent alteration decreases

the stability of communication services and increases

the cost of system operation and maintenance.

To prevent the alteration of switching networks,

off-switch solutions use Intelligent Network (IN) as

the implementation basis (Lin & Rao 1999). An

NPDB is implemented for maintaining portable

numbers and the corresponding destination addresses.

Here we assume a global NPDB (GDB) which is

centrally maintained by a neutral organization

(Number Portability Administration Center, NPAC)

is available. All portable numbers of every operator

are recorded in GDB. Operators may make a copy of

NPDB to their networks to omit the delay of long

term GDB.

There are four off-switch schemes for solving the

NP problem: all call query (ACQ), query on release

(QOR), call dropback (also known as return to pivot,

RTP), and onward routing (OR) (Kim & Yong 2003).

The four schemes exist in the NP solutions of

different countries: UK, Finland, France, Germany,

Span, Singapore, etc. <http://www.arcome.com>,

<http://www.ida.gov.sg>, <http://www.tct.hut.fi>.

The considerations of which scheme to adopt include

the network resources, the policy of addressing and

routing, the impacts on the signaling system, and the

interworking with other services. ACQ and call

dropback are the two most popular solutions.

z All call query (ACQ): The originating network

initiates a query to NPAC when determining the

dialed number is a portable number. NPAC

returns the destination address of the number,

then the originating network routes the call to the

destination address to set up the call. It is the

most efficient of using the network transmission

facilities.

Call dropback: The originating network receives

a call and routes the call to the donor network by

the prefix of the number. The donor network

determines the number was ported out, it returns

the routing address of the number’s new

subscription network and release the call. Then

the origination network re-routes the call to the

destination network. The routing information is

maintained in the donor networks of portable

numbers. However, transmission resources are

occupied during the routing of calls. This may

reduce the efficiency of source network.

The drawback of the existing methods results

from the growing database size. Every call

terminated to a portable number initiates a database

query. Numerous NP service subscribers incur a

cumbersome database that severely encumbers the

efficiency of routing information query and call

setup process.

3 APPLYING CACHES TO PSTN

3.1 Analysis of calling behavior

Cache performs well when the data variation is

infrequent and the data has access locality.

Implementing caches to switching centers the size of

the cache either is too large to be practical or the hit

rate is too low to be efficacious. On the other hand,

caches on terminals suffer from the alteration of

subscription networks, which will cause the change

of caches in all terminals.

In fix-lined telecommunication systems, an

organization (e.g., a company, an enterprise, a

campus, or a government department) often

establishes PBX for saving telecommunication cost.

A PBX is an exchange which enables organization

members to call each other freely, and to make and

receive calls from users served by the public network.

ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES

138

The communications within a PBX service region

are routed by the PBX, without using the

communication resource beyond. A PBX has one or

more physical channels that connect to a local

exchange, the communications beyond the PBX are

routed to the public network through the local

exchange (Black 1998).

The communication targets of an organization

usually have locality. For example, organizations

like government departments or retail businesses

have frequent contact objects to solve problems or to

supply merchandise. The communication targets of

an organization may be diffused, but always include

a set of cooperation subjects. From the view of

telecommunication network hierarchy, the quantity

of users within a PBX service region is moderate,

and the locality of PBX contact targets is

distinguishable.

Portable numbers are determined as external calls

of PBX that is routed to local exchanges. The local

exchange determines whether the request can be

processed locally, or should be routed to a higher

layer switch. If a switch can translate a portable

number, it consults NPDB to determine the routing

information of the number to routes the call;

otherwise, it routes the request to the higher layer

switches for number translation.

3.2 Applying caches to PBX

3.2.1 The operation of caches in the fix-lined

telecommunication system

For maintenance and long-term operation, operators

generally adopt off-switch schemes for NP

communication services. The purpose of caches on

PBX is to share the workload of portable number

translation, and to alleviate the traffic load of NPDB

for shortening the call setup time and enhancing the

utilization of communication resources.

Caches on PBX keep the routing information of

dialed numbers (both portable and non-portable

numbers) to omit the delay of NPDB queries. A

mechanism is needed for PBX to inform the

confirmed routing information of the called parties

to the public switching system, thus the call can be

routed directly without querying NPDB.

The communication beyond PBX must through

the public communication system. Conventionally,

PBX route every external call to the public network

through local exchanges. When caches are applied to

PBX, a PBX checks the internal cache before routing

an external call. A cache hit indicates that the

destination address of the dialed number is

confirmed, and the NPDB queries can be omitted.

PBX adds a special code (e.g., *14*, *30*) in front

of the dialed number to indicate the destination

address of a dialed number was confirmed and

appended to the call originating request. The public

switching system recognizes the code and routes the

call by the appended routing information without

querying GDB or donor NPDB.

PBX

cache

Exchange

NPDB

Terminating

switch

Subscription

GW/NPAC

NPDB

Ben

*14*Ben_dest

Ben_dest

Ben

NPDB

Originating network Terminating network

Donor/NPAC network

Figure 1: A special code is used to indicate a cache hit

Figure 1 illustrates the basic NP call process. A

call is made from a PBX to an external user Ben.

PBX determines the request is an external call and

checks cache for the destination address of Ben (step

1). In the case of cache hit, PBX appends a special

code “*14*” and the obtained destination address to

the call originating request. The request is sent to a

local exchange (step 2). The public switching system

routes the call to the called party Ben without

querying NPDB (step 5 & 6).

When a cache miss, the routing information of

the called party should be solved by higher level

exchanges. The PBX routes the call originating

request to the local exchange without modifying the

request or adding any special code. The public

switching system determines that is a NP call, and

consults NPDB to translate the number (step 3 & 4),

and routes the call to the called party (step 5 & 6).

The signaling systems of operators usually have

the function to process special codes (e.g., the code

“*67” in USA prohibits displaying the caller’s

telephone number to the called party). Applying

caches to PBX requires adding a rule for new special

code in the signaling system without altering the

operation logic. It is feasible for operators to

implement the new service.

3.2.2 Issues of applying caches to PBX

Following the communication model of PBX with

caches, the issues of cache implementation

encompass the policy of establishing a cache, the

update of cached data, and the size of the cache.

z The policy of cache establishment

Caches should keep as many as possible the

destination addresses of portable numbers, but the

size of a cache is restricted that the data can be kept

in is limited. The cache hit rate increases when the

A PBX-BASED APPROACH FOR TELECOMMUNICATION NUMBER PORTABILITY SERVICE

139

accessed data has locality. To improve the cache hit

ratio to enhance the efficiency of NP call setup, the

cache needs to expose the communication habits of

users of the PBX service region.

There are two approaches for cache

establishment: The dynamic cache policy supposes

the most recently dialed numbers shall be cached for

they might be dialed repeatedly in a span; from

another point, the static cache policy believes the

frequently dialed numbers must be cached for they

represent the most frequently contact targets for a

long-range observation.

When the subscription network of a cached

number changed, the cached routing information

becomes obsolete. This will bring about miss-routing

calls that consume extra signaling and transmission

resources and prolongs the call setup time. Follow

the dynamic cache policy, the altered routing

information must be updated dynamically, thus the

signaling system of the public network must be

modified to notify PBX to renew the routing

information immediately. Besides, cache

replacement happens whenever a new data is

inserted to the cache; this may replace expected data

and decline the hit rate. The costs to modify the

signaling system of public networks, to management

a dynamic cache, and to update routing information

in caches are expensive. Consequently, dynamic

cache policy is not feasible for PBX caches to

enhance the performance of NP service.

On the contrary, a static cache keeps the most

frequently dialed numbers (FDN) of the PBX. The

establishment and alteration of a static cache is

manually performed by the system administrator of

an organization. When a member joined an

organization, the member proposed a set of FDN to

the system administrator. The system administrator

sends the FDN set to a contracted operator to obtain

the corresponding routing information. This is a

batch process that can prevent inconsistent data in

the cache. When a contact target changes

subscription network, the contracted operator sends

messages to registered PBX to notify the alteration.

While the contact targets of an organization have

locality, and the change of members is infrequent,

the variation of FDN is gentle. The complexity of

establishing and maintaining a static cache is much

less than that of a dynamic cache. The hit rate of a

static cache is proportional to the probability the

FDN be dialed.

z Cache update

The cache update can be initiated by PBX or by

operators. Operators can provide a service for PBX

to consult the routing information, and the

consulting can be processed off-line in off-time to

reduce the impact on operators and PBX, even can

be served through Internet. Or a PBX can register a

profile of FDN to the contracted operator, thus the

operator can notify the PBX to renew altered routing

information.

NPAC

New subscription

Switch

GDB

cache

PBX subscription network New subscription network

NPAC network

NPDB

PBX

Subscription

switch

NPDB

Old

subscription network

profile

Figure 2: The process of cache update

Assuming a neutral GDB is available, the update

of caches is illustrated in Figure 2. When a portable

number changes subscription network, the new

subscription network updates the new destination

address of the number to NPAC (step 1 & 2). NPAC

notifies the alteration to contracted operators (step 3).

When received a notification of altered routing

information, the subscription network of PBX

checks registered profiles (step 4), and sends a

notification to PBX (step 6). Then PBX initiates a

query process to query the altered information, and

updates the new data to the cache (step 7).

For limiting the porting frequency of NP

subscribers, we assume users can not change

subscription networks too often. The effectiveness of

a new ported number is postponed for couple hours

to guarantee the consistency of GDB and the NPDB

of old and new subscription networks.

z Cache size

A cache entry consists of a 10- to 15-digit

telephone numbers which follows ITU-T

Recommendation E.164 (Faltstrom 2000), a 4-octet

destination address of the number, a 4-byte field of

the latest modification time, and a 2-bit cache

management tag. A cache entry occupies 186 bits.

Assume that there are s members in an

organization, the FDN quota of each member is k.

Suppose the probability the FDN of a member

overlap with other members’ is r in average, 0≤ r <1.

Let U be the universal set of all individual FDN set

of member i, which can be represented as

() ( )

UUU

kji

uuuuuuU

≠=≠===

∩∩+∩−=

,0,,,0,00

≠

jikjijijii

(

)

uuu

∩

The minimum cache size ms

cache

to keep the FDN set

of an organization is

cache

)1(

−

∑

⎟

⎠

⎞

⎜

⎝

⎛

−

ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES

140

∑

, where 0<r< 1. (1)

−−

⎟

⎠

⎞

⎜

⎝

⎛

−×

)1(

When the contact targets of every member is

scattered that r = 0, the minimum size of the cache

of the iNetwork is represented as

cache

= s × k, where r = 0. (2)

In an organization with 15000 employees, each

member has 100 FDN in average, when r = 0, the

total cache size is less than 33.3 MB; when r = 0.5,

the cache size reduced to 17 MB. It is obvious that

caches are small enough to be located in memory

that the search time is less than 0.01 msec memory

access. Since a GDB query consumes 2 seconds or

less, when GDB queries are substituted by memory

accesses, the delay of querying routing information

can be reduced remarkably.

4 PERFORMANCE ANALYSIS

PBX with a cache diminishes the demand of

long-term database queries, and decrease the NP call

setup time. But it also requires the modification of

PBX. In this section the cost and the benefit of PBX

with caches are evaluated.

The costs of applying caches to PBX includes the

cost to modify the signaling system, the cost to

establish caches and update cached data, and the

time to query cached data.

The signaling system needs to add a new rule on

switches to process special codes. PBX must be

modified to query caches before forwarding requests,

and to append the destination address and a special

code to a call initiation message. These are the

constant investment of system construction.

The cost of cache implementation includes the

hardware cost of memory, the cost of the batch

process to attain the routing information of FDN,

and the cost to enter FDN manually. While memory

is cheap, the entering of FDN is infrequent, and the

query of routing information of FDN is an off-line

batch process that can be executed in off-time, the

different of conventional NP service systems and the

system with caches on PBX is the cost of system

operation. We study the benefit of PBX with caches

by comparing the cost of call initiation process of the

conventional telecom system and the cache-applied

system.

When caches are applied to PBX, every external

call triggers a cache query. The result of a cache

query can be a cache miss when the required data is

not kept in the cache, or a location hit if the cached

data points to the actual destination address of the

dialed number. Location miss will not occurred

because of the postponed activity of NP subscribers.

The cost of call setup is represented as

call setup with cache

where

and

are the cost of call setup when it is

a cache miss and a location hit, respectively;

and

are the probability of cache miss and location hit

respectively,

=1.

includes the costs of cache

access and conventional NP/non-NP call setup

process,

equals to the probability of dialing

non-FDN.

= C

cache access

+ C

donor/NPAC NPDB query

+ C

message processing

+ C

service

message transmission

+ C

queuing

= P{dial non-FDN} = 1-P{dial FDN}

A location hit is the most economic case of call setup.

is represented as

= C

cache access

+ C

message processing

+ C

service

+ C

message transmission

queuing

= P {dial FDN} = 1 -

The message transfer delay of a trunk is denoted as

trans

. Database access time includes the time for

GDB query (t

GDB

), and for local NPDB query (t

NPDB

The time for cache query is t

cache

. The total delay

time of cache access, message process and

transmission, service, and queuing is denoted as t

Different NP schemes results in the diverseness

of network operation and signal flows. We study the

performance of PBX caches by analyzing the call

setup delay of ACQ and call dropback.

Case 1: ACQ

z In the conventional NP call process, the call

setup time t

acq_conv

can be represented as

acq_conv

= t

trans

+( t

GDB

+ t

NPDB

)+ t

(3)

z In the case of location hits, the call setup time

acq_hit

can be represented as

acq_hit

trans

+ ( t

cache

+ t

NPDB

)+t

(4)

= (3) + t

cache

– t

GDB

Where t

cache

is much less than t

GDB

that can be

ignored, and (4) is much less than (3).

z In the cache miss case, the call setup time is:

acq_miss

= t

trans

+(t

cache

GDB

= (3) + t

NPDB

)+t

(5)

cache

Case 2: Call dropback

z The call setup time t

cd_conv

is represented as

cd_conv

= t

trans

+(2×t

NPDB

) + t

(6)

z In the case of location hits, the call setup time

cd_hit

can be represented as

cd_hit

= t

trans

+( t

cache

+ t

NPDB

)+ t

(7)

= (6) + t

cache

– t

NPDB

z When cache miss, the call setup time t

cd_miss

trans

+(t

cache

NPDB

)+ t

= (6) + t

(8)

cache

Assuming the call arrival rate in a PBX is a

Poisson process at the rate of λ per second, the

average call setup time in a PBX is

Total call setup time

conventional

t × t

conventional

A PBX-BASED APPROACH FOR TELECOMMUNICATION NUMBER PORTABILITY SERVICE

141

In the case of PBX with caches, the probability of

using FDN is p in average, the call setup time is

Total call setup time

cache

t ×( p × t

hit

+ (1-p) × t

miss

)

From equations (5) and (8), the conventional call

setup time closes to that of cache miss (p= 0).

Assuming the longest connection is less than 450

km. The time to process a request needs 100msec.

The time to query GDB data should be less than 2

sec (Foster, McGarry & Yu 2003), the time for

NPDB query should be less than 350 msec (Gans,

King & Woodbridge 2001). A PBX handles 2000

calls per hour. The delay of searching a cached entry

is less than 1/100 millisecond in average. According

to the increase of p, the total NP call setup time with

caches decreases linearly, as shown in Figure 3.

Average call setup time

500

1000

1500

2000

2500

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

t(in msec)

cached ACQ

cached dropback

ACQ

Dropback

Figure 3 : The relation of FDN utility rate (p) and the

average call setup time (t in msec)

SCP access times (p=0.7)

100

150

200

250

300

350

400

450

500

50 100 150 200 250 300 350 400 450

SCP access_cache

SCP access_no cache

Figure 4: The SCP access times per second

The query delay dominates the NP call setup time.

Numerous NP users results in substantial NPDB

queries, NPDB access becomes the bottleneck of NP

call process. Take ACQ as an example, if caches are

not available, or the cache queries are all missed, the

setup time of a call is 1911 milliseconds in average.

When 70% of the dialed numbers are FDN, the call

setup time decreased to 1211 msec. It reduced to

almost half of that without caches.

Calls originated from organization dominate the

telecommunication traffic in business hours.

Assuming 70% of the calls are originated from

enterprises in business hours, and 85% of enterprises

has established PBX. The number of NPDB accesses

per second is demoted as a. According to the

increase of a, the use of FDN cache can reduce the

DB load effectively (Figure 4).

5 CONCLUSION

In this article we stated the importance of NP

services in telecommunication systems, and the

problem of long information query delay resulted

from huge NPDB. By analyzing the hierarchy of

fix-lined telecom network, we proposed to apply

caches to PBX to share the traffic load of NPDB,

and to alleviate the delay time of querying

destination addresses of portable numbers.

According to the communication characteristics of

organization members, we suggest adopting static

cache policy, which is easy to establish and maintain.

Caches on PBX alleviate workload of NPDB,

and remarkably reduce the average NP call setup

time. The improved NP call setup efficiency leads to

better communication resource utilization; therefore,

caches on PBX benefit both the telecommunication

users and service providers.

This solution needs not to modify the signaling

system of the public telecommunication network.

Hence, applying caches to PBX is a low-cost

solution to enhance the performance of the NP

service in the fixed-line telecommunication system.

REFERENCES

Gans J., King S., Woodbridge G., Numbers to the people:

regulation, ownership and local number portability,

Information Economics and Policy Vol. 13, Issue 2,

June, 2001, pp. 167-180.

Lin Y. B, Rao H., Number portability for

telecommunication networks, IEEE Network, Vol. 13

Issue 1, Jan.-Feb. 1999, pp. 56 –62.

Foster M., McGarry T., Yu J., Number portability in the

Global switched telephone network (GSTN): An

overview, RFC 3482, Feb. 2003.

Black U., The intelligent network, customizing

telecommunication networks and services, Prentice

Hall, 1998.

Carpenter T., Carter R., Cochinwala M., and Eiger M.,

Data Caching for Telephony services, International

Database Engineering and Application Symposium,

2000, Sep. 2000, pp. 101-109.

Faltstrom P., E.164 number and DNS, RFC 2916, Sep.

2000.

Kim H. S., Yong H. S., Personalized cache management

for mobile computing environments, Information

Processing Letters Vol. 87, Issue 4, Aug. 2003, pp.

221-228.

ICETE 2005 - GLOBAL COMMUNICATION INFORMATION SYSTEMS AND SERVICES

142