Capacity of an Alcatel 1000 S12 exchange emphasised on the ISDN remote subscriber unit

BY JOHN YTTERHAUG, GUNNAR NOSSUM AND ROLV-ERIK SPILLING

1 General
This article describes the dimensioning of the main equipment of an Alcatel 1000 S12 exchange and the corresponding traffic restrictions where this is appropriate. It is divided into two main parts: The first part describes the Alcatel 1000 S12 exchange, and the second part describes the remote concentrator called ISDN Remote Subscriber Unit (IRSU). The IRSU is developed and maintained in Norway. The Alcatel 1000 S12 is a common Alcatel product with main development in Belgium. Each country may have different requirements for the equipment, reliability and grade of service. This article is based on the dimensioning rules for the Norwegian market. Most of the Alcatel 1000 S12 exchanges in Norway have old software. The existing hardware in an exchange can with some adaptations be used with the new software package including ISDN, Business Communication, etc. Full functionality is therefore possible for all exchanges. All types of modules needed for this can be introduced, either in existing racks or by introduction of new rack types. The description here only concerns the latest versions of the modules and rack types.

planes are identically equipped. The number of switching units in each plane is determined by the number of TSUs.

capacity, however, will be reduced to the half. No hard restrictions exist on the penetration figures for subscriber facilities. Extending the requirements for capacity and availability is always possible by proper dimensioning of the required resources. It is possible for all subscrib-

2.2 Terminal Sub Units (TSU)

A TSU consists of one group of terminal modules. One TSU is connected to one Access Switch (AS) pair. Figure 2 shows how two TSUs are connected to two Access Switches. The figure only shows the processors (TCE is Terminal Control Element) of the terminal modules. Each AS is connected to all four switching planes in the DSN. The number of TSUs is calculated based on the total traffic demand and the total amount of terminal equipment. The maximum allowed mean traffic per TSU (determined by the GOS requirement for end-to-end blocking probability in the switching network) for 4 planes is 152 Erlang. This ensures that the signalling messages between the modules have a high probability to succeed. If a free path is not found, several reattempts are done, the number of reattempts depending on the importance of the message. The number of subscriber modules and other traffic dependent modules per TSU is calculated based on the TSU traffic constraints given above.

ASM ISM IRIM CCM HCCM SCM

T C E T C E T C E T C E T C E T C E

T C E T C E T C E T C E T C E T C E T C E

TTM DTM IPTM APM MPM CTM DIAM

Digital Switching Network DSN

S A C E

2 Dimensioning of common exchange equipment

This part describes the main equipment in an Alcatel 1000 S12 exchange and the capacities and dimensioning of the equipment where appropriate.

Figure 1

Access Switch (AS)

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 9 10 11 12 13 14 15

2.3 Subscriber modules

Two types of subscriber modules exist, one for analogue and one for digital (ISDN) subscribers. The maximum traffic per subscriber module is 35.2 Erlang. The subscriber module is connected to the DSN via two separate PCM systems. With a GOS of 0.2 % the traffic capacity is 17.6 Erlang per connection. If the required traffic per line gives a higher total traffic per module than 35.2 Erlang, the subscriber modules will be underequipped. Modules for analogue and digital subscribers may be combined in the same TSU. 2 modules with the same type of subscribers work as cross-over pairs, which means that if one module fails another will take over the function. The
T S U TCE 1 TCE 8

2.1 Digital Switching Network (DSN)

An exchange is built up around a Digital Switching Network (DSN). The DSN transports circuit switched, packet switched and signalling traffic. All the modules in the system are connected to the DSN as shown in Figure 1. The modules are connected via Terminal Sub Units (TSU). Two access switches are equipped per TSU, and they connect the modules to the switching planes in the DSN. The following modules in Figure 1 are not described later in the text: TTM: Trunk Testing Module CTM: Clock and Tone Module. The number of switching planes is 3 or 4 depending on the traffic. All switching

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 8 9 10 11 12 13 14 15

DSN

T S U

TCE 1 TCE 8

Figure 2

87

ers to use the same facility at the same time provided the dimensioning has been done accordingly. One subscriber can have a combination of facilities and can use his facilities in the same call as far as they are not incompatible. The influence of combinations on the dimensioning can be neglected. 2.3.1 Analogue Subscriber Module (ASM) The module terminates 128 analogue lines, with a maximum traffic of 0.275 Erlang per line in a fully equipped ASM. An ASM consists of two common PBAs (processor and ring PBA) and from 1 to 8 subscriber PBAs with 16 analogue lines per PBA. 2.3.2 ISDN Subscriber Module (ISM) The module terminates 64 Basic Access (BA) lines, with a maximum traffic of 0.550 Erlang per line in a fully equipped ISM. An ISM consists of one common PBA (processor) and from 1 to 8 subscriber PBAs with 8 ISDN lines per PBA.

2.5.1 Digital Echo Cancelling Module (DEC) One module with integrated echo cancellation equipment terminates one first order PCM link with different versions for No. 5 signalling and No. 7 signalling. One module consists of two PBAs (one processor and one echo cancelling PBA). 2.5.2 ISDN Packet Trunk Module (IPTM) This module has several applications. The hardware is the same, but the software varies. Each of the IPTM modules consists of two PBAs (one processor and one trunk PBA). - Primary Rate Access (PRA) This module can be used for termination of an ISDN PABX connected with a PRA connection. - Common Channel Signalling (CCS) This module has a lower capacity per link than the HCCM module, and is more cost efficient if only a few No. 7 links are required. The module terminates up to 4 No. 7 channels in a PCM link. The rest of the channels can be used as traffic channels. The capacity of the signalling terminal under normal load conditions is 400 messages per second per module. - Frame Handler (FH) The frame handler part comprises concentration and multiplexing of packet switched traffic. This module terminates up to 4 packet channels in a PCM link. The rest of the channels can be used as traffic channels. The twoway throughput capacities under normal load conditions is 300 frames per second per module. - X25 This module terminates up to 4 packet channels in a PCM link. - Packet Handler Interface (PHI) This module is used as a gateway function for the Packet Handler in DATAPAK. 2.5.3 Digital Trunk Module (DTM) One module terminates one first order PCM link with channel associated signalling or transparent No. 7 signalling. The module consists of only one PBA.

2.5.4 ISDN Remote Interface Module (IRIM) One IRIM/S, respectively one IRIM/T, is used for termination of 1, respectively 2, links to a singledrop (SD) IRSU or a multidrop (MD) of IRSUs. 2 IRIMs work as a cross-over pair. The module consists of two PBAs (one processor and one trunk PBA, DTRF for double trunk and DTRH for single trunk).

2.6 Dynamic Integrated Announcement Module (DIAM)

The new announcement module DIAM will provide the following types of announcements: - Fixed announcements (e.g. Call Forwarding) - Combined fixed and variable announcements (e.g. Changed Number Interception) - Variable announcements (e.g. Personalised Changed Number). Two versions exist: - DIAM-basic, one PBA version. Requires 4 Mbyte storage capacity on disk. Speech storage capacity is 8 minutes - DIAM-extended, two PBAs version. Requires additional 20 Mbyte of storage capacity on disk (a total of 24 Mbyte for the whole module). Speech storage capacity is 50 minutes. The number of speech channels are: - 56 channels can be used for speech - 4 channels are reserved for virtual path communication. The maximum duration for one announcement is flexible and only limited by the speech store capacity. If more announcement capacity is needed per exchange, more modules may be equipped.

2.4 High Performance Common Channel Signalling Module (HCCM)

The exchange has several possibilities for terminating No. 7 links. The most common is the use of the HCCM module which controls up to 8 external No. 7 links. This module consists of one processor PBA and one line PBA per No. 7 link. The signalling terminal is capable of receiving and sending N7 messages with only one flag between the messages. The two-way throughput capacity under normal load conditions is 500 messages per second per link. The transmission capacity is 64 kbit/s in each direction. With a message length of 20 bytes this will give a load of 0.625 Erlang on the No. 7 link. With normal holding times and utilisation of the 2 Mbit/s one No. 7 link can signal for several hundred 2 Mbit/s with todays type of traffic.

2.7 Service Circuit Module (SCM)

The modules are dimensioned in separate groups per signalling type. Within each group a cyclic selection mechanism with 4 trials is used to find an SCM with free service circuits. For reliability reasons only 3 trials is assumed for dimensioning purposes. Groups containing R2 or R5 signalling units are dimensioned according to the

2.5 Trunk modules

Different types of trunk modules are provided and they are described in the following sections.

88

GOS requirement of 0.1 % waiting probability for seizing of a free circuit. For exchanges without MFC signalling, the GOS requirement for DTMF/R2 signalling units will be 1.0 %. The modules will be equipped in different modes: - One MF-unit with 16 senders/receivers for R2/DTMF and one Simplified Conference Bridge (SCB). The SCB can handle 6 simultaneous conferences with maximum 5 participants per conference. The number of SCB units is calculated based on the originating traffic using the services Add on Conference and Meet-me Conference and a part of the originating traffic using the services Three Party Service, Call Hold and Call Transfer. The GOS requirement is 1.0 %. - One MF-unit for R2/DTMF and one service unit for Continuity Check Transceivers (CCT). CCT equipment is used for No. 7 traffic on international routes. One CCT service unit contains 16 CCT devices. - Two MF-units for R2/DTMF with a total of 32 senders/receivers. - One MF-unit for R5 and one service unit for CCT (Continuity Check Transceivers). Each of these SCM types will consist of only one PBA in addition to the processor. As a general rule, an extra module of any equipped type will be added for reliability reasons if the number of modules of that type is less than or equal to 4.

- PABX lines: 10 % of all analogue subscriber lines and basic accesses - Traffic per PABX line: 0.4 Erlang. In the following sections the SACEs are described. 2.9.1 Call Service SACE This SACE contains the following functions: - Prefix analysis and task element definition - Local subscriber identification - Charging analysis - Call control - Charge generation control - Trunk request co-ordination - Facility processing - Alarm call processing. The Call Service SACE works in load sharing mode for all functions. Therefore, the Call service SACE has to be split into load sharing groups. With service penetration figures used for Norway, a load sharing group has a capacity of 9600 subscribers. For security reasons all data are replicated over at least two processors, which are working in load sharing for processing the data accesses. Each load sharing group is equipped with one extra processor if the number of processors calculated is less than 5. 2.9.2 PBX/CHRG SACE This SACE contains the functions to manage PBX lines (analogue and digital), Centrex hunt groups and the charging function. PBX/CHRG SACE contains the following functions: Automatic message accounting Meter counts collection Local tax layout Private access resource management BCG resource management Call control for PABXs Charge generation control for PABXs.

required for interfacing a taxation centre. One pair working in active/standby mode is equipped. 2.9.4 IN/OSI SACE This SACE contains IN call control and OSI-stack modules working in load sharing mode. An extra processor is equipped if the required number of processors is less than 4. 2.9.5 IDC/TRA SACE This SACE contains the following functions: Line intermediate data collection Trunk intermediate data collection Trunk request allocation Device interworking data collection.

The processor load is calculated for each function. 5 % of the processor capacity is reserved for storage of measurement counters on disk, assuming a storage interval of 15 minutes. One extra processor is always equipped.

2.10 Network Services Centres (NSC)

If the exchange has a collocated NSC, an Administration and Peripheral Module (APM) has to be added. This module contains functions for administration and peripherals in addition to the specific NSC-function. These two functions are not split into separate processors. The APM works together with magnetic and optical disk. Two modules are equipped per NSC and are the same as the MPMs.

2.8 Maintenance and Peripheral Module (MPM)

This module contains functions for peripherals and loading and works together with the magnetic and optical disk. This module also contains functions for administration. Two modules are equipped per exchange.

2.11 Taxation Centre

A Taxation Centre consists of a Transaction Processor (TP) and a Billing Processor (BP). The TP is an integrated part of the exchange. Each TP can handle 64 exchanges or 100,000 subscribers. The TP is a duplicated processor system where the Administration and Peripheral Modules (APM) work in active/active mode.

2.9 System Auxiliary Control Elements (SACE)

A minimum number of SACEs are required per exchange. For new SW releases the installed processors are reused. If not explicitly given, the following traffic is assumed: - Centrex lines: 5 % of all analogue subscriber lines and basic accesses - Traffic per Centrex line: 0.2 Erlang

3 ISDN Remote Subscriber Unit (IRSU)

The purpose of the IRSU is to provide a cheap subscriber connection to an Alcatel 1000 S12 exchange. As the name Remote Subscriber Unit indicates, the use is primarily for clusters of remote subscribers relative to the exchange. The advantage

The PBX/CHRG works in active/standby. 2.9.3 CCSN7/DEF SACE This SACE contains the defence, No. 7 and operation and maintenance functions. In addition, this SACE contains functions

89

IRIM with cluster links

29-m 30-n

M-IRSU

IRSU

JR05 is an outdoor cabinet which has space for a JR01 rack, transmission equipment, termination for optical fibre cables, power supply and batteries.

CE
29-m

DTRF
30-n

Multidrop links
29-m 30-n

3.3 Dimensioning of trunks in a multidrop

Dimensioning of trunks is done based on a GOS requirement of 0.5 % for the circuit switched (CS) traffic between the IRSU and the group switch in the exchange. The GOS requirement is for the worst case subscriber. In the special configurations described later, the difference of GOS for subscribers can be substantial. The capacities given later are valid for configurations with one packet switched (PS) channel per MD link. It is possible to allocate one PS channel on one MD link and none on another, but the traffic capacities for such configurations are not described. The allocation of a PS channel replaces a CS channel and vice versa. Thus, as the number of allocated PS channels increases, the CS traffic capacity is reduced. The two main dimensioning parts of the system is the MD links which each has 30 channels and the cluster side links (see Figure 3) of the IRIM which has 29 channels each for circuit and PS traffic. The number of CS channels is shown in Figure 3. In this figure, m and n is the number of PS channels on the cluster links and MD links, respectively. For a DTRH configuration in the IRIM, n = 1 will imply m = 1. For a DTRF configuration in the IRIM, n = 1 will imply m = 2. The IRSU has the same schematic structure as the IRIM. Instead of the Control Elements (CE, i.e. the processor) there are clock and alarm PBAs. Each subscriber PBA is connected to one direct and one cross-over link between the clock and alarm PBA and the DTRF/DTRH. To fully understand the traffic capacities in the following sections, it is important to note that the traffic from a subscriber group on the upper side has to go through the upper processor (control element) in the IRIM and similarly for the lower side subscribers. The normal channel allocation algorithm is first to try the direct route, i.e. from a subscriber on the upper to the upper control element in the IRIM through the direct cluster link. Only if this direct route is blocked, the cross-over cluster link in the IRIM will be tried. The same

CE
29-m

DTRF
30-n

Figure 3

of introducing an IRSU is a considerable saving on cables if the alternative is to connect each of the remote subscribers to the exchange itself. If the alternative is to put a new exchange on the remote site, the solution of an IRSU is much cheaper and will save administration and maintenance costs since the administration of the IRSU is done from the host exchange. Only HW maintenance is necessary on the site of the IRSU. An IRSU is connected to the exchange through an IRIM pair. Up to eight IRSUs may be connected in a series and is then called a multidrop (MD). One IRSU connected to an IRIM pair is called a singledrop (SD). An SD or an MD is connected to one IRIM pair. An MD is used primarily for low traffic subscribers, typically a fully equipped MD can carry up to 0.083 Erlang per equivalent line. An analogue line and a digital line are one and two equivalent lines respectively. It may also be used for high traffic subscribers, but underequipping is then necessary. It is important to note that an IRSU does not have a processor PBA that takes care of the call handling which is handled in the exchange in the same way as for directly connected subscribers.

- The IRIM/T is equipped with a DTRF trunk PBA which terminates two PCM links (double trunk system). On the IRSU side the DTRH will usually be connected to an IRIM/S, while the DTRF will be connected to an IRIM/T. An IRIM has only one processor. A maximum of 4 MD links can be assigned to one MD configuration (2 x IRIM/T). The upper and lower side is cross-connected by cross-over links in the IRSUs as well as in the IRIM modules except for one trunk PBA configurations (such as mini IRSU and self restrained configurations). This assures that from each subscriber all trunks can be reached. Available standard types of PBAs are used for connecting the subscribers. Between indicated maximum limits, any 16 analogue subscribers can be replaced by 8 ISDN subscribers.

3.2 IRSU variants

Each variant consist of one subrack containing one (JR01), two (JR02) or three shelves (JR03). A common rack (JR00) has space for six shelves grouped three and three. The JR04 rack (mini IRSU) will occupy one shelf in the JR00 rack. The transmission equipment for one IRSU is provided by a separate subrack (JR06). The shelf variants have the following subscriber capacities: - JR01: 256 ASL + 0 BA to 32 ASL + 112 BA - JR02: 512 ASL + 0 BA to 32 ASL + 240 BA - JR03: 976 ASL + 24 BA to 64 ASL + 480 BA - JR04: 96 ASL + 0 BA to 0 ASL + 48 BA.

3.1 Multidrop description

An MD consists of two half-systems, called the upper and lower side. A maximum of 8 IRSUs can be connected to the same MD system with a maximum of 1024 equivalent lines. A maximum of 4 subscriber groups of 128 equivalent lines may be assigned to each IRIM in the IRIM pair. - The IRIM/S is equipped with a DTRH trunk PBA which terminates one PCM link (single trunk system).

90

procedure is followed for subscribers on the lower. 3.3.1 Normal configurations Traffic limits for the normal trunk configurations, which are the two link and four link configurations, are as follows: Two DTRH PBAs: Two DTRF PBAs: 43.0 E 82.5 E

the IRSUs. The IRSUs have subscribers on both upper and lower side. Traffic limits for the one PBA configurations are as follows: One DTRH PBA: One DTRF PBA: 18.2 E 34.8 E

In the case of two DTRH PBAs the MD links are the bottleneck, i.e. Erlangs formula for 58 channels is used. In the case of two DTRF PBAs the direct and crossover cluster link in the IRIM are the bottleneck, i.e. Erlangs formula for 56 channels multiplied by two is used. These limits are based on the use of the same trunk PBAs in the IRSUs in the MD, i.e. that all IRSUs have 2 DTRHs or 2 DTRFs, respectively. If several IRSUs with DTRHs are connected to a DTRF in the IRIM, the DTRH PBAs should be distributed equally on the two MD links for capacity reasons. Furthermore, it is assumed that each IRSU has subscribers on both the upper and lower side of the system, i.e. at least two subscriber groups per IRSU. In an MD configuration, the total traffic from all IRSUs has to be within the above given limits. 3.3.2 Special configurations If the traffic from the IRSUs is much smaller than 43.0 Erlang or a little more than 43.0 Erlang, some special configurations may be considered. These configurations are normally not recommended due to reliability reasons or traffic reasons (because of an asymmetric configuration). 3.3.2.1 One trunk PBA configurations The economical one trunk PBA configuration can be used if the traffic is low and the reliability aspect is not considered that important. Figure 4 shows a one trunk PBA configuration with DTRF PBA in the IRIM and

In the case of DTRH PBA the MD link is the bottleneck, i.e. Erlangs formula for 29 channels is used. Very unbalanced traffic is well tolerated. For example there can be 17.0 Erlang from subscribers on the upper side and 1.2 Erlang from subscribers on the lower side and vice versa. In the case of DTRF PBA the bottleneck is the direct and cross-over cluster link in the IRIM, i.e. Erlangs formula for 28 channels multiplied by two is used. It must here be assumed that the traffic on the upper and lower is well balanced. If the trunk PBA in the IRIM or the trunk PBA in the IRSU fails, the connection between the IRIM and the IRSU will be disrupted, whereas for a normal configuration the traffic will use the alternative cross-over link when the direct link is disrupted. 3.3.3 Asymmetric configuration This configuration has one DTRF PBA and one DTRH PBA in the IRIM and in the IRSU as well, i.e. three MD links. Figure 5 shows an asymmetric configuration. The asymmetric configuration may seem a good solution for traffic needs between the traffic capacity for a configuration with two DTRHs and two DTRFs in the IRIM. But it should only be used with the alternative channel allocation algorithm which chooses the MD link pair with most free channels. Without this alternative channel allocation algorithm the capacity can be improved by placing more subscribers on the side (upper or lower) where the DTRH is placed, which at first may seem surprising. When the traffic on the DTRH side is twice the traffic on the DTRF side the traffic capacity is 61

Erlang. In this case the high traffic from the DTRH will be blocked on the cluster link, but different from the opposite traffic situation the overflow traffic together with the direct traffic will now have two MD links available on the DTRF side. There will be very little overflow traffic from the DTRF side to the DTRH side. The point is to distribute the traffic so that the cluster links carry equal traffic as much as possible. If the alternative channel allocation algorithm is used the performance of the balanced upper and lower traffic case is improved. This algorithm has the effect of distributing the traffic more evenly and thus reducing the more bursty overflow traffic. With balanced traffic on the upper and lower side the traffic capacity is 66.3 Erlang and is dimensioned according to Erlangs formulae with 87 channels minus PS channels. So, if this alternative channel allocation algorithm is used the asymmetric configuration may be considered.

3.4 Mini IRSU

The mini IRSU is an inexpensive, simple IRSU. The mini IRSU is special in two respects. It contains only one subscriber group and it has only one trunk PBA (DTRH or DTRF) which means it has no cross-over link. This means that a mini IRSU subscriber is a worst case subscriber when the mini IRSU is in an MD with other non mini IRSUs. To fulfil the GOS requirement for all subscribers in an MD it is necessary to decrease the total traffic if the mini IRSU traffic increases. The traffic capacity for the mini IRSU is 17.4 Erlang. The bottleneck is the cluster link, hence Erlangs formula with 28 channels is used. However, the bottleneck is quickly changed to the MD link with 29 channels when other non mini IRSUs are connected to the same MD and there are DTRHs in the IRIM. If there are DTRFs in the IRIM the bottleneck is again the cluster link. Figure 6 shows a configuration with DTRH PBAs in the IRIM and the IRSUs

Figure 4

Figure 5

Figure 6

91

and one mini IRSU and one normal IRSU in the MD. The mini IRSU is an economical way to introduce ISDN subscribers if there are IRSUs connected to the exchange already. The rest of this section is based on an alternative channel allocation algorithm which chooses a channel on the MD link pair with most free channels. This has the effect of distributing the traffic more evenly on upper and lower MD links. If one mini IRSU is placed on the upper and one is placed on the lower, the total capacity for the MD is 34.8 Erlang regardless of which trunk PBAs that are equipped and whether there are other than mini IRSUs or not in the MD. Thus, it is better to place all the mini IRSUs on one side (if there are fewer than five mini IRSUs). This is particularly important in the case of DTRFs in the IRIM. If this is done, the total traffic capacity for the MD depends on the traffic from the mini IRSUs and the more the traffic from the mini IRSUs approaches the limit of 17.4 Erlang the less traffic the MD as a whole can carry. Since there is no bursty overflow traffic that mixes with normal traffic it is easy to find analytical models. In the following formulae this notation is used: Am_u: Traffic from mini IRSU which is connected to upper side Au: Al: Lc: Lm: Traffic from upper side (excluding mini IRSU traffic) Traffic from lower side Traffic limit on one cluster link Traffic limit on one MD link

Am u +

Au2 Lm (Au + Am u)

capacity falls continuously as the mini IRSU traffic increases for the DTRF case.

If there is 10.0 Erlang coming from the mini IRSU then Au + Al must be less than 28.1 Erlang. Simulations show that the limits given by this formula is conservative for all values of the traffic from the mini IRSU. This is mainly due to less burstiness than ordinary Poisson traffic for the traffic on the MD links from the ordinary IRSUs. A secondary effect is that high mini IRSU traffic will push the other traffic over to the other MD link which will help the throughput of mini IRSU traffic. For dimensioning purposes the interesting traffic interval for the mini IRSU is between 5 and 15 Erlang. As a uniform rule in this interval the total MD capacity could be set to two times the traffic capacity on one MD link, i.e. 36.4 Erlang even though a higher traffic can be carried. If there are DTRFs in the IRIM the direct cluster link on the side of the mini IRSU is the restricting part of the system for all distributions of Au and Al. The following formula is given for the case of two DTRFs in the IRIM:
Am u +
Au+Al 2 Au+Al + Am 2

3.5 Dimensioning of PS channels in a multidrop

As mentioned in the Dimensioning of trunks in an MD section the total number of circuit and PS channels is constant. The nature of PS and CS switched traffic is very different. The CS traffic is a socalled lost calls cleared system whereas the PS traffic is a so-called waiting call system. The latter system does not clear the call if there are no recourses available at the moment. Each packet of a PS call is treated according to a waiting call system. The different nature of the two types of strategy for packet and CS traffic can be exploited by the network operator. It could be tempting to minimise the number of PS channels. However, the delays for the packets cannot be allowed to be too long because of buffer size and protocols. If a packet is delayed more than a certain amount of time the packet is retransmitted and when many packets are retransmitted the traffic increases which in turn gives even longer waiting times. 3.5.1 The PS system Some characteristics for the PS system are listed below: - Receive and send side in the IRIM and IRSU work in parallel. - Each PS channel is used only in one direction, from the IRIM through IRSU 1 to the highest numbered IRSU and back to the IRIM. - Packets sent from the IRIM is received by one of the IRSUs and sent on the PS channel through the loop to the IRIM where it is absorbed. Packets sent from an IRSU is received and absorbed by the IRIM. - Each IRSU and the IRIM are only allowed to send one packet at a time before possibly being interrupted by another IRSU or the IRIM. - Each IRSU and the IRIM have one queue per PS channel. 3.5.2 Analytical model To describe the system correctly there should exist one queue in every drop which experiences arrivals of PS calls both from its own subscribers and the

Au Lc 2 u

If Au = Al the formula can be reduced to:

Am u + Au2 Lc (Au + Am u) 2

If there are DTRHs in the IRIM the MD link on the side of the mini IRSU is the restricting part of the system for most distributions of Au and Al. Only if Al is far higher than Au (under the assumption that the mini IRSU is placed on the upper side) will the restricting part of the system be moved from the MD link to the upper direct cluster link. The following formula applies assuming Au and Al are approximately equal:
Am u +
Au+Al 2 Au+Al + Am 2

The second part of the sum is the traffic from the normal IRSUs offered to the direct upper cluster link of the IRIM. So, for example for 10.0 Erlang from the mini IRSU, the total traffic for a two DTRF PBAs configured MD is 53.3 Erlang assuming that the traffic from the other IRSUs are evenly distributed on upper and lower. Simulations show that the traffic limits derived from the formulae above is conservative for all values of Am_u and more so the higher mini IRSU traffic. The total traffic capacity is reduced as the mini IRSU traffic increases. Thus, the difference between the DTRH case and the DTRF case is that the total traffic capacity is more or less constant up to a certain mini IRSU traffic for the DTRH case, whereas the total traffic

Au + Al Lm 2

If Au = Al the formula can be reduced to:

92

previous drops in the MD and from the IRIM (incoming traffic). There are assumed n M/M/1 queuing systems (Markovian input process, Markovian service time distribution, one server) with first in first out (FIFO) queue discipline per PS channel where n is the number of PS channels. The n channels serve the whole MD. The server time is assumed to be proportional to the packet length with a service speed of 64 kbit/s. Two traffic models are given for the PS traffic, the medium traffic and the high traffic model. The medium and high traffic model has 10 and 50 call attempts per hour (CAPH) per subscriber respectively. Each call has an average of 18 packets with an average length of 125 bytes. Since both outgoing and incoming PS calls are queued successively in all IRSUs and the IRSUs are considered as one common queue, all calls are considered outgoing. The maximum number of ISDN subscribers in an MD is 512. The results in the following sections are based on this number of subscribers. It is important to note that the tables reflect the total PS traffic equivalent to 512 subscribers and that realistic PS traffic will be much lower due to fewer ISDN subscribers in an MD. 3.5.3 Results In the following text and tables the following abbreviations occur: : : CH: CHT: CAPH: DELq: Packet arrival rate per second for a MD Service rate in packets per channel Number of PS channels Call Holding Time Call Attempts Per Hour Average delay in milliseconds (ms) for a delayed packet

From Table 1 can be seen that for medium PS traffic one or two packet channels are sufficient. For high PS traffic three or four channels are sufficient, but one or two channels is impossible since the load is 2.0 and 1.0 respectively (a load higher than 1.0 gives an infinite queue). The delays depend on the packet length given the same load. In Table 2 the bytes per packet and packets per call are interchanged so that there are many short packets instead of few long packets. Short packets reduce the delays and average queue lengths. Since an objective is to minimise the number of long delays to avoid retransmission, shorter packets may be used. 3.5.4 PS traffic in the signalling channel The PS traffic intensity generated by ISDN subscribers may be so small that an allocation of one separate PS channel, and hence a reduction in CS traffic capacity, would be an unsatisfactory utilisation of the system. Thus, the signalling (S) channel may be used as a combined channel for PS traffic and S traffic. This solution utilises the spare capacity of the S channel for PS traffic. The same queuing model as in the section Analytical model has been used. The abbreviation Wpacket is used for the weighted average packet length in bytes based on the distribution of CS and PS packets. The 2 CH and 4 CH columns in the tables correspond to a DTRH and a DTRF configuration respectively. CAPH and CHT is based on 0.08 Erlang per subscriber. CSpackets/call is 88 packets on an average, i.e. 88 S packets per CS call. CSbytes/packet is 33 bytes on an average. This assumes that all traffic carried in the MD originates and terminates here, which is a worst case. Approximately 2/3 signalling is for the originating side. 1/3 is for incoming traffic. Since only few calls use maximal signalling and the rest is balanced, the theoretical results should be slightly conservative. In Table 3 only S traffic is considered. The results in Table 3 clearly show that even with high S traffic the load contribution due to S traffic will be modest (worst case load-DTRH = 24.8 %). The average waiting time for all packets will not exceed 1.35 ms, and 95 % of the

Table 1 PS Traffic
CAPH/SUB Bytes/packet Packets/call 1 CH Load DELq DELall Qbyte 95%_del 95 %_all 25.6 0.40 26.0 10.4 33.3 78 54 10 125 18 25.6 64 2 CH 12.8 0.20 19.5 3.9 6.3 59 27 3 CH 42.7 0.67 46.9 31.3 166.7 140 121 50 125 18 128.0 64 4 CH 32.0 0.50 31.3 15.6 62.5 94 72

Table 2 PS Traffic
CAPH/SUB Bytes/packet Packets/call 1 CH Load DELq DELall Qbyte 95%_del 95%_all 177.8 0.40 3.8 1.5 4.8 11 8 10 18 125 444 2 CH 88.9 0.20 2.8 0.6 0.9 8 4 50 18 125 444 3 CH 296.3 0.67 6.8 4.5 24.0 20 18 4 CH 222.2 0.50 4.5 2.2 9.0 14 10

Table 3 Signalling Traffic

CAPH/SUB cht Bytes/packet Packets/call 2 CH Load DELq DELall Qbyte 95%_del 0.082 4.4 0.37 0.25 13 3.2 90 33 88 40.04 242.42 4 CH 0.041 4.3 0.18 0.06 12 2 CH 0.248 5.4 1.35 2.69 16 9.6 30 33 88 120.16 242.42 4 CH 0.124 4.7 0.50 0.58 14

DELall: Average delay in ms for all packets Qbyte: Load: Average number of queued bytes per channel Load per channel

95%_all: Delay in ms for which 95 % of all packets are below 95%_del: Delay in ms for which 95 % of the delayed packets are below.

93

Table 4 Combined signalling and packet traffic

CSpackets/call PSpackets/call CSbytes/packet PSbytes/packet 2 CH CScaph/sub PScaph/sub Total caph/sub Bytes/Wpacket (Wpacket/s) (Wpacket/s) Load DELq DELall Qbyte 95%_del 3.2 10 13.2 68.9 65.6 116.2 0.28 12.0 3.4 7.7 36 4 CH 3.2 50 53.2 103.1 168.0 77.6 0.54 28.1 15.2 65.8 84 88 18 33 125 2 CH 9.6 10 19.6 49.2 145.7 162.7 0.45 11.1 5.0 17.9 33 4 CH 9.6 50 59.6 80.4 248.1 99.4 0.62 26.7 16.7 83.3 80

4 Summary
The dimensioning of an Alcatel S12 exchange is done according to the description in chapter 2. The dimensioning is done in a very flexible way. A traffic increase, whether it comes from (new) ASMs or DTMs can easily be handled by increasing the number of SACEs, SCMs and other equipment. The IRSUs in an MD can be configured in many ways as described in chapter 3. For example can one large IRSU or up to 8 smaller IRSUs be connected to an IRIM pair. In addition, the number of links (2 Mbit/s) connected to the IRIMpair can vary, but the recommended number of links are two or four. In addition, a mini IRSU may be connected, but care must be taken because of special traffic limitations. The D channel used by digital subscribers connected to an MD is normally connected to allocated PS channels on the MD links. One PS channel per link may be assigned. If the PS traffic is low, there will be a possibility to use the signalling channel both for signalling and packet traffic which means that the circuit traffic capacity is increased.

delayed packets will not experience a waiting time longer than 16 ms. Combining S and PS traffic through the parameter Wpacket, higher waiting times can be expected. Table 4 shows the effect of mixing PS traffic and S traffic, see also Table 1 which has the same PS traffic. The highest 95%_del has the DTRF configuration with 50 CAPH for the PS traffic and 90 seconds CHT (3.2 CAPH) for CS traffic. But the highest DEL_all is seen for the DTRF configuration with the same PS traffic and 30 seconds CHT for CS traffic (higher load). The loads given in the tables do not give serious delays, i.e. delays that cause a lot of retransmissions. The link access protocol has a threshold of 1000 ms before a retransmission occurs and thus retransmission should be a rare incident. The model is very simple so with more bursty traffic, delays and queues may be longer than indicated in the tables, but since the margins are high the risk of jamming down the system with retransmissions should be minimal.

References
For a thorough, but not up to date description of Alcatel 1000 S12 the following references are recommended: 1 2 ITT 1240 Digital Exchange, Electrical communication, 56(2/3), 1981. System 12 Digital Exchange, Electrical Communication, 59(1/2), 1985.

For a more thorough view on the Digital Switching Network and general dimensioning the following reference is useful: 3 Haukeland, J. Hvordan dimensjonere en System 12 sentral. 9th Nordic teletraffic seminar (NTS 9). Kjeller 1990.

94