NO 4,1995

The Introduction of UniSwttch in AXE 10 Copper Enhancement MINI LINK E - A New Link for Flexible Transmission in Cellular Networks Dynamic Routing in Circuit-Switched Networks Ericsson Review - 70 Years Young

CONTENTS No. 4 1995 • Vol.72

The Introduction of UniSwitch in AXE 10

145

Copper Enhancement

153

MINI-LINK E - A New Link for Flexible Transmission in Cellular Networks

160

Dynamic Routing in

Circuit-Switched Networks

169

Ericsson Review - 70 Years Young

178

Cover: The MINI-LINK E microwave system - providing improved flexibility for meeting transmission requirements in many different applications and networks.

Photos: Karl-Evert Eklund Kent Eliasson Progema

CONTENTS Previous issues No. 4 , 1 9 9 4 ATM Traffic Management at the Initial Deployment of B-ISDN Re-defining Management Systems The TMOS Architecture Evolution Trends in Wide Area Paging Network Traffic Management (NTM) Using AXE and TMOS Systems MOMS - An Operations System for MINI-LINK No. 1, 1 9 9 5 APZ 21220 - The New High-end Processor Measuring Quality of Service in Public Telecommunications Network AXE 10 System Processing Capacity

Using Predictions to Improve Software Reliability Test Marketing of Mobile Intelligent Network Services AXE 10 Dependability

Development of AXE for New and Very Demanding Transit/Tandem Switching Application No. 3 , 1 9 9 5 System T - A Modular System for Wide Area Paging

No. 2 , 1 9 9 5 A Swedish Airborne Early Warning System Based on the Ericsson Erieye Radar High-performance Packaging for a RISC Processor Application D-AMPS 1900 - The Dual-Band Personal Communications System

Using DECT for Radio in the Local Loop A System for Flexible Service-Independent Access Network Solutions Handling Overload in AXE 10 UniSwitch - A New Flexible STM Switch Fabric Concept

The Apollo Demonstrator - New LowCost Technologies for Optical Interconnects

Ericsson Review © Telefonaktiebolaget L M Ericsson • Stockholm 1995. • Responsible publisher Hakan Jansson • Editor Steve Banner • Editorial staff Eva Karlstein • Layout Paues Media • Address Telefonaktiebolaget L M Ericsson S-126 25 Stockholm, Sweden • Fax +46 8 681 27 10 .Published in English and Spanish with four issues per year • Subscription one year USD 45 Ericsson Review No. 4. 1995

143

CONTRIBUTORS in this issue

Stefan Gustafsson, senior systems engineer for switching network products in Ericsson Telecom's core unit Basic Systems, works on system coordination and early investigations in the group switch area. At present, he is system manager of the project launched for the development of UniSwitch for AXE 10. Giinther Hell, manager of the Local Loop Transmission competence centre of Ericsson-Schrack, is responsible forthe development, management and sales of products in this area. After passing his engineer's examination, he has been working in the telecommunications business for twelve years; the six latest years with Ericsson-Schrack, holding his present position since 1992. Andy Rolfe, sales and marketing manager at Ericsson Communications Ltd, Local Network Systems division, New Zealand, is responsible for marketing access network products in Europe. Before joining the company, he spent more than eight years with Ericsson Ltd, England: six years in engineering and two years as access account manager. Kent G Ahlqvist, area manager at the Microwave Communications division of Ericsson Microwave Systems, is responsible for the marketing of MINI-LINK in a number of European countries. Hans Andersson works as a specialist in mathematical statistics in the traffic theory group of the systems management department of Ericsson Telecom's core unit Basic Systems. He earned his MSc in Electrical Engineering from the Royal Institute of Technology in 1970. Mattias Lindeberg works as a systems engineer in the systems management department of Ericsson Telecom's core unit Basic Systems. He earned his MSc in Electrical Engineering from the Lund Institute of Technology in 1994.

144

Ericsson Review No. 4, 1995

The Introduction of UniSwitch in AXE 10 Stefan Gustafsson

The group switch in AXE 1 0 - 64K GSS - has been continuously improved since its release more than a decade ago, which means that it is still abreast of the times. It functions very well and satisfies most of customers' needs. But forthcoming demands and requirements can only be met through major redesign. That is why the UniSwitch will be introduced in AXE 1 0 , as the new - future-proof - group switch. The author describes the improvements and possibilities that will be effected through the introduction of the GS-128K UniSwitch.

The characteristics and possibilities of the UniSwitch concept, as described in no. 3 / 1 9 9 5 of Ericsson Review, are those of a truly future-proof system. UniSwitch - together with new versions of the APZ control system - is capable of meeting increasingly greater demands on the core of AXE: APZ and the GSS. Also, the continuous updating of No. 7 signalling and the adaptation of SDHbased transport networks and IN-based architectures to the system give promise of bright prospects of unbroken success for AXE 10. These are the reasons why the UniSwitch has been chosen as a hardware platform for the next generation of group switches for AXE 10.

eriangs for a fully equipped GS-128K switch. Non-blocking The UniSwitch is strictly non-blocking i.e. zero congestion1. This characteristic is one important improvement com-

Fig. l Subrack containing one plane of a hilly equipped 1 2 8 K UniSwitch core.

Improvements compared with the existing group switch The existing group switch has a maximum capacity of 64K inlets/outlets (channels); the UniSwitch has twice this capacity, 128K. Switching is performed at the 64 kbit/s level. 64K channels will hardly suffice for future demands. Future networks will use optical fibre to transport enormous volumes of traffic in and between megacities of up to 30 million inhabitants. Today's switch will have to be redesigned to meet these extreme capacity requirements.2 The first version of the UniSwitch to be introduced in AXE 10 is the GS-128K, with a capacity of 128K 64 kbit/s channels for traffic. Another 12K 64 kbit/s channels are available for signalling and connection of auxiliary equipment. Each UniSwitch channel will be capable of handling 1.0 eriangs (no internal blocking). The total capacity figure for switched traffic is typically 50,000 Ericsson Review No. 4, 1995

145

Fig. 2 The relationship between a fully equipped 6 4 K switch and UniSwitch, with adaptations to existing equipment and with upgraded equipment without any need tor adaptations. Equal size: 6 4 K inlets.

pared with the existing 64K switch. Nonblocking allows any combination of inlets/outlets to be used for switched or leased line connections, with bit rates between 64 kbit/s and 155 Mbit/s without any restrictions. Digital cross-connect functionality at the 1/0 level is also possible.

Box A Abbreviations ATM CACB CP CT CTCM CTIB DLI DL2 DXC ECB EMB ET

146

Asynchronous transfer mode Cable connection board Central processor Control terminal Control terminal connection module Control terminal interface board Digital link interface Digital link 2 Digital cross-connect Exchange clock board Extension module bus Exchange terminal

Footprint The need for a smaller and smaller footprint has been a particular objective each time the existing group switch has been updated. This has led to a reduction in physical size. As appears from Fig. 2, the introduction of the GS-128K will reduce the switch size even more.

GSS ICB NSCM RP RPB SDH STM-1 TCB TCU TSSI TU TUCM USCM USI

Group switching subsystem Incoming clock board Network synchronisation connection module Regional processor Regional processor bus Synchronous digital hierarchy SDH at 155 Mbit/s Terminal unit connection board Terminal connection unit Time slot sequence integrity Terminal unit Terminal unit connection module UniSwitch core module UniSwitch interface

Ericsson Review No. 4 , 1 9 9 5

The GS-128K switch core is housed in three 600 mm deep BYB 5 0 1 cabinets. The control terminals for communication with the central processor and some network synchronisation equipment are accommodated in one BYB 202 cabinet. Terminal connection units, TCUs, are needed to connect lower transmission rates, e.g. 2 Mbit/s terminals. The particular type of TCU that performs multiplexing/demultiplexing from these 2 Mbit/s terminals is placed in thirteen BYB 501 cabinets, in a 128K configuration. One of the cabinets also contains a subrack with AXE 10 network synchronisation equipment for the UniSwitch. This reduces the size of a 64K inlet switch to one third of that of the existing group switch, Fig. 2. Optimised 24-channel connection Connection of 24 channel units to the UniSwitch is accomplished without using any special equipment to transform four 24channel PCM systems into three 32-channel PCM systems. This is possible thanks to the mapping performed in the terminal connection units when equipment is connected to them. The purpose of the mapping function is to meet the variable bandwidth requirements of different types of equipment (in this case the need for only

24 channels) while efficiently utilising the switch core. The mapping function can handle any number of channels; not only 24 or 32.

rig. 3 Time slot sequence integrity, TSSI. A, B, C, D, E and F are single 64 kbit/s channels on one wideband connection. TSSI means that after the channels have passed the switch, they are in the same order: A, B, C, D, E, F.

Wideband without restrictions A new requirement that has to be met by today's network operators is the demand for wideband: several ordinary 64 kbit/s channels grouped together for different purposes. Time slot sequence integrity, TSSI, must be guaranteed for the channels employed, Fig. 3. One wideband application is videoconferencing, which needs 6 x 64 kbit/s for acceptable quality to be achieved. The UniSwitch hardware supports wideband connections without restrictions. A

Fig. 4 A control terminal connection module magazine equipped with one control terminal. Ericsson Review No. 4, 1995

147

Fig. 5 UniSwitch in A X E 1 0 , with adaptation equipment.

user of this function can order the switch to set up any number of channels, between 2 and 32 (2 Mbit/s), as one wideband connection and even - potentially - STM-oriented broadband connections up to 155 Mbit/s. A-law or u-law idle pattern In some markets, e.g. the US, there is a demand for different idle patterns for different terminating equipment. The existing group switch can send only one type of idle pattern. Conversion to other patterns must be performed outside the switch. In the UniSwitch, different idle patterns are supported by the hardware. For each set of terminating equipment, one pattern - A-law or u.-law - can be selected. New

148

patterns can be introduced without hardware modification. Power consumption The power consumption of a UniSwich with 64K inlets is reduced to one fourth of that of the present 64K GSS. Reliability Triplication of the hardware makes the UniSwitch single-fault tolerant, allowing greater maintenance intervals-for example, weekly scheduled maintenance of the switch hardware. Fast circuit switching The fast circuit switching in the UniSwitch also allows packet-oriented switching. Ericsson Review No. 4, 1995

Fig. 6 A network synchronisation connection module magazine with board positions for twenty external synchronisation references and three exchange clock boards.

This means that the UniSwitch may also serve as a medium for interprocessor signalling: the system is capable of application-specific processor support, e.g. No. 7 signalling. DXC applications The fact that the UniSwitch hardware allows integration of digital cross-connect capabilities can be of great advantage in nodes where separate DXC equipment is not justified. ATM transport node The switch interface - USI4 - is dimensioned for a bit rate exceeding 155 Mbit/s, which means that the UniSwitch is ideal for the introduction of Ericsson Review No. 4. 1995

transmission terminations, such as STM-1 in AXE 10. It can also serve as a transport access node for broadband services to users in the AXE 10 exchange service area; for example, accessing a centrally located ATM broadband switch.

Hardware units in the AXE 10 application of UniSwitch Fig. 5 shows an overall picture of the different UniSwitch units - briefly described in the following. See no. 3/1995 of Ericsson Review for more details of the concept. 149

Fig. 7 A T U C M 2 x 4 subrack with 3 2 positions for TUs, e.g. DLIB and the three T C U boards.

UniSwitch core module, USCM The core module performs the actual switching procedures. A USC128K module is used in the first application for AXE 10. The USCMs are always placed in three double-depth cabinets. Terminal connection unit, TCU The terminal connection unit, TCU, performs the multiplexing/demultiplexing and mapping of channels in the UniSwitch. The first variant to be used in the AXE 10 application is the TCU2x4; it will include terminal connection unit boards (TCB2x4) placed in a terminal unit connection module (TUCM2x4), see Fig. 7. This subrack connects 32 terminal units, TU, with USI2 interfaces. The normal configuration is two cascade-connected TUCM2x4 subracks for 2 Mbit/s terminat-

Box B

Clock regulation

The three exchange clock boards (ECB) contain the clock generators that are used to stabilise the internal regulation performed in the UniSwitch core. The clock generators receive input both from the UniSwitch core and from external sources via incoming clock board 1 , ICB1. One of the ECBs is master clock; the others are controlled by it. In case of a master ECB failure, one of the other ECBs takes over as the master clock.

150

ing equipment. Two TUCM2x4 subracks can handle 2,048 channels, which fit into one USI4 inlet to the USCM. In the first UniSwitch application, each of the TU positions in the subrack is occupied by a digital link interface, DLI, board. The DLI is the logical name of the interface unit that converts the existing electrical interface, DL2, to the new electrical interface, USI2. Implementation of USI2 in future terminating equipment, e.g. new exchange terminals, ET, eliminates the need for DLI as an adaptation unit. This will reduce the footprint even more, compared with the existing switch, Fig. 2. Control terminal, CT The control interface functions between the central processor, CP, and the UniSwitch are performed by control termi-

One hundred times per second, the master ECB sends values of the phase difference between the switch core and the master ECB, to ensure the network synchronisation function. Each ECB is controlled by a microprocessor, which reads the regulator values, calculates the differences and ensures that the differing values are sent to the core regulators. The ECBs are connected to the switch via front connectors on the TCB2x4 board or via the CACB.

Ericsson Review No. 4, 1995

Fig. 8 Exchange clock board for the network synchronisation connection module.

nals, CT, located in a subrack designated CTCM (control terminal connection module). Up to four CTs can be housed in one CTCM. A CT consists of two parts: the regional processor type device, RPD, and the control terminal interface board 2, CTIB2. The RPD is one of the existing regional processors in AXE 10 3 . There is no EM bus interface in the RPD; instead, a processor bus interface is available in the backplane of the subrack. The CTIB2 is the interface between the RPD and the UniSwitch. It connects to the backplane bus of the RPD and converts the signals to the USI2 interface. A microprocessor controls this conversion. The CTIB2 is connected to the switch via front connectors in the TUCM subrack, or via the cable connection board, CACB. The CACB can be placed in any of the TU positions in the TUCM subrack. Only sixteen CTs - placed in four CTCMs - are required to control the operation of a fully equipped 128K UniSwitch.

Incoming clock board 1 , ICB1 The incoming clock board 1, ICB1, converts the different external network synchronisation sources to one electrical interface that is used in the NSCM subrack. External sources may be signals from other exchanges or local synchronisation sources, e.g. an atomic clock or some other reliable source. Up to twenty boards can be placed in the subrack, which means that twenty different sources can be connected.

Network synchronisation connection module, NSCM The network synchronisation connection module, NSCM, contains the different sets of synchronisation and timing equipment that are required to adapt the UniSwitch to the AXE 10 environment. One subrack contains all the synchronisation interface equipment needed for the AXE 10 application.

One important issue in the development of software for the UniSwitch has been that of keeping the existing software interfaces to the switch unchanged, if possible. Interfaces for the connection of units to the switch had to be modified, and the new hardware structure also necessitated some modifications. But, by and large, these modifications have only affected one subsystem outside the GSS:

Ericsson Review No. 4, 1995

Exchange clock board, ECB Three exchange clock boards, ECB, are placed in the NSCM subrack. This triplication resembles the successful arrangement in the existing 64K switch, which means that the principles of clock regulation presently applied can be reused.

Software for the UniSwitch

151

contains some functions common to the existing group switch and UniSwitch. This solution makes it possible to keep the well-proven 64K software unchanged, in cases where it must be available along with the new UniSwitch software. Both PLEX and C are used as programming languages for the different software units. The device processor, DP, used in the UniSwitch is loadable via the APZ control system. This makes for easy updates of the software when new functionality is needed in the system.

Future development of the UniSwitch concept in AXE 10

Fig. 9 Software structures of the new GSS releases: The GSS manager, common functions, and the packages for the control of different hardware units.

the operation and maintenance subsystem, OMS. Minor modifications have been made in the extended switch subsystem, ESS, and in some other subsystems. For ordinary users of the group switch, the software interface is unchanged. The new software should also allow control of the existing switch hardware; i.e. the customer should be able to keep the same software application system regardless of switch hardware. This requirement has been met by the introduction of the GSS manager, Fig. 9, which is the software interface to the switch. Three software packages belong under the manager. One package is the existing 64K software; the second is the UniSwitch software package, and the third

References Lundh, P. and Roos, S.: UniSwitch - A New Flexible STM Switch-Fabric Concept, Ericsson Review 74 (1995):3, pp. 132-142. Moberg, L. and Sandberg, K.: Development of AXE for New and Very Demanding Transit/Tandem Switching Applications, Ericsson Review 74 (1995):2, pp. 89-98. Hjalmarson, T.: AXE 10 Control System, Ericsson Review 69 (1990):3, pp. 119-129.

152

The need for a large, 128K switch is not self-evident for mobile and localexchange applications, although the switch core is scalable from 8K up to maximum size, in steps of 8K. The next switch version has already been planned, to meet the need for a sizeoptimised UniSwitch. Today's implementation makes it possible to build all the three planes of a 32K switch into one subrack. This subrack will be placed in a single-depth BYB 5 0 1 cabinet. The Uniswitch concept allows the switch core to be even more reduced in size. A 16K switch core on three printed board assemblies will be possible in the future. And, to shift the perspective, the concept will also enable the building of switch cores larger than 140K.

Summary The first version of UniSwitch to be introduced in AXE 10 - GS-128K - is a complement to the present 64K GSS; not a replacement. The switching capacity of the 64K GSS is perfectly adequate for many applications today and tomorrow. But the introduction of UniSwitch means substantially increased capability to offer new facilities; for example, wideband connections without restrictions. It will also enable customers to build large metropolitan exchanges with small footprints. UniSwitch is a flexible platform, designed to provide present functionality along with new applications - such as SDH termination and DXC capabilities - which is of vital importance to truly future-proof systems. Ericsson Review No. 4, 1995

Copper Enhancement Gunther Hell and Andy Rolfe

Copper pairs - used in the local loop - account for more than 90% of the six hundred million telephone connections to public switched telephone networks (PSTN) in the world. The fact that this number is expected to grow to nine hundred million by the year 2 0 0 0 makes further exploitation of the existing copper infrastructure well worth considering. The authors describe the traditional use of copper in the local loop and discuss some new and emerging technologies that are or ought to be used to enhance copper pairs.

New technologies and new regulatory conditions are continuously changing the economics of the local loop. Telecommunications in general, and lately voice telephony in particular, are being liberalised and - as a result - "access" in the local network has become a major issue. New access operators use fibre optic networks when they connect large customers, focusing on the high revenues this segment generates. For an emerging but largely unknown base of minor, lowrevenue customers, access is provided through radio in the local loop (RLL). Traditional telecom operators (TOs), on their part, need to reassess their present ownership of the local loop and are increasingly looking for new ways to exploit the existing infrastructure. The technological development behind the change in the local loop is not confined to traditional telecommunications. Multimedia and interactive enter-(edu-)tainment, the Internet, the soaring processing power of personal computers, video games, CD-ROM, microchip technology, software etc. are all driving towards a requirement for bandwidth to become a commodity in global networking.

Transmission media in the local loop The majority of TOs have an extensive copper access network. When these networks were initially deployed, voice telephony was virtually the only service provided - naturally so, because hardly any other service was required. Therefore, the local loop only consisted of cabling structures with one twisted pair as the basic element. The capacity of cables (i.e. the number of copper pairs per cable) decreased as they approached Ericsson Review No. 4, 1995

the subscriber, which necessitated the use of distribution points (cross-connects) in the "outside plant", in a treelike architecture. More recently, TOs have started to use other technologies - such as fibre and radio - to connect subscribers. Fibre is economical in the large business customer segment, as is radio in the low-density residential segment. However, technologies continue to develop and the economic scenario changes as a consequence. In spite of these competing technologies, copper pairs account for more than 90% of the six hundred million PSTN connections worldwide. It is therefore important to look at their role in the changing regulatory environment, where more operators will have to gain access to subscribers. The number of PSTN lines is estimated to grow to nine hundred million by the year 2000, Fig. 1, and demand for new services (and their associated bandwidth) is also expected to continue to grow. This provides more than sufficient reason for maximising the use of the existing copper infrastructure. The picture will be dramatically changed if we include cable TV into the

Box A Abbreviations ADPCM Adaptive differential pulse code modulation ADSL Asymmetric digital subscriber line ANSI American National Standards Institute BER Bit error rate CAP Carrierless amplitude modulation DMT Discrete multi-tone ETSI European Telecommunications Standards Institute FTTB Fibre to the building

Fig. 1 Number of main lines installed worldwide in 1 9 8 0 and 1 9 9 0 and an estimate for 2 0 0 0 according to I T U (the International Telecommunication Union).

FTTC HDSL HFC ISDN LAN MPEG PGS POTS PSTN RIBS VDSL VoD

Fibre to the curb High-bit-rate digital subscriber line Hybrid fibre coax Integrated services digital network Local area network Moving picture experts group Pair gain system Plain old telephone system Public switched telephone network Residential interactive broadband services Very-high-bit-rate digital subscriber line Video on demand

153

\

Fig. 2 Relationship between the maximum speed of access network transmission systems and the maximum distance to be achieved at this speed. (Over short distances, very high speed can be provided on normal copper pairs.)

considerations regarding the local loop. Today, approximately one hundred million homes are connected to cable TV; in other words, 16-17% of all PSTN connections. The local copper plant will have to be exploited further, so that we can consider it a valuable asset rather than a liability for the traditional TOs.

The capacity of the local copper-pair cable The copper pairs traditionally used in the local network - one pair of copper wires connecting one telephone subscriber burdened the monopolistic PTOs with heavy investments in outside plant engineering, notably cable installation. Some early technologies already provided the means to use one pair for more than one voice telephony connection, and party line systems were quite popular for some time. They provided access to the PSTN, albeit not with 100% availability. Subscribers had to physically share one copper pair, which meant that while one subscriber was using the telephone, the others were denied access. Voice-band and baseband modems, 2 Mbit/s transmission systems in the 1970s and 1980s and, later, ISDN demonstrated that a copper pair could trans-

port much more than one voice-band channel at a time. Special cables, capable of transporting one or more such channels, were designed in order to accommodate 2 Mbit/s transmission. Once technology had allowed the capacity of baseband transmission to grow to several hundred kilobits per second, and the distances between repeaters to go beyond twice the average local loop distance, the obvious next target was to reach 2 Mbit/s and even more. The graph in Fig. 2 illustrates the current view of the capacity of twisted pair cables in the local loop. Table 1 gives an overview of various technologies, with the distances and transmission speeds they can provide.

Transmission technologies ISDN In 1980, CCITT set up a standard defining the frame conditions for a worldwide digital network called ISDN. The aim was twofold: to have international, standardised interfaces to subscribers, usingdigital transmission technology; and the economical provision of high-quality services. As set out in the CCITT standard, transmission of the ISDN basic rate access (BRA) has a bit rate of 160 kbit/s using different line codes, e.g. 2B1Q, 4B3T and TCM. The standard also defines the ISDN primary rate access (PRA) for transmission of up to 30 channels. BRA delivers ISDN services to the subscriber over a standard twisted pair loop. The digital loop carries two B-channels for voice or data with a bit rate of up to 64 kbit/s, and one 16 kbit/s D-channel, which handles call-control messaging for the B-channels as well as packet data. Customer premises equipment, such as ISDN telephones, data terminals, PCs

Table 1 Overview of existing and emerging technologies with the distances and transmission speeds they can provide

154

Technology

Km

kbit/s

VDSL very-high-bit-rate digital subscriber line

0.4

50,000

ADSL asymmetric digital subscriber line

4.5

8,000

HDSL high-bit-rate digital subscriber line

8

2,048

ISDN integrated services digital network

10

160

Baseband modem

25

9.6

Ericsson Review No. 4, 1995

Fig. 3 Typical pair gain system configuration with the PGS exchange board (EU) installed next to the local exchange and the remote unit (RU) installed next to the customer premises using one standard telephone cable.

and fax machines, is connected to the BRA. PRA delivers wider bandwidth to systems with ISDN capabilities, such as digital PBXs, host computers and LANs. It carries 23 or 30 B-channels, each with a bit rate of 64 kbit/s, and one 64 kbit/s D-channel for messaging.

Digital pair gain systems (PGS) The introduction of ISDN and the development of chips to transmit 160 kbit/s on twisted copper pairs, using the 2B1Q line code, formed the basis of the development of first generation PGS, such as PCM-2 (also called 0+2) and PCM-4. PCM-2 allows two subscribers (PCM-4 up to four) to be connected to the switch, over one twisted copper pair. PGS characteristics Pair gain systems provide simultaneous transmission of 2-12 telephone connections over one twisted copper pair. This means that they multiply the capacity of existing access networks without increasing the number of lines (cables). Digital transmission technology achieves optimum performance and maximum distances besides making the systems immune to interference. The main advantages of PGS are: - multiple use of copper pairs to increase the number of subscribers - economical and efficient solution - digital transmission technology (ISDN or HDSL) - ease of installation and maintenance - remote power feeding - fast response to customer requests for more telephone lines Ericsson Review No. 4, 1995

Technology The 2B1Q transceiver chip is also used for transmission in PCM-2 and PCM-4. Today, 50% of these chips - produced worldwide - are in fact used for PGS. With a 2B1Q chip, a network operator can connect up to four subscribers per system, over a single twisted pair. The most recent developments in PGS will use HDSL chips with speeds of 784 or 1,168 kbit/s to allow more than four subscribers over a single twisted pair.

HDSL, High-bit-rate digital subscriber line HDSL is a transmission technology that will provide high-speed digital access on non-loaded telephone loop plant in a repeaterless subscriber-loop network. The speed is typically 2 Mbit/s. HDSL enables high bandwidth to be delivered in both directions (duplex). HDSL characteristics - transmission over existing unconditioned, twisted copper pairs - duplex rates of up to 2 Mbit/s - repeaterless operation - no pair selection required - very low bit error rate (BER) - line code: 2B1Q Technology HDSL has been developed in the US to carry 784 kbit/s duplex on a single copper pair over a distance of around 3.5 km. This allows it to carry a US T l signal (1.5 Mbit/s) over two pairs. However, the European standard uses a 2 Mbit/s signal, which means that there are two H DSL options in Europe: - use of three pairs with 784 kbit/s (US

155

Fig. 4 Typical HDSL configuration with the HDSL exchange board (EU) installed next to the local exchange and the remote unit (RU) installed next to the customer premises using two standard telephone cables.

standard) for the 2 Mbit/s European standard - use of two pairs with a bit rate of 1,168 kbit/s per pair The standard line code is usually 2B1Q but ETSITM3 has now also accepted the description of CAP (carrierless amplitude modulation) as an Annex. Since 2B1Q with a CAP Annex is approved as an ETSI Technical Report, it will be up to the market to decide which one will prevail. Lately, European TOs have indicated that the two-pair solution, using the 2B1Q code, is preferable. This is mainly because the cable network structure is based on one- or two-pair applications. HDSL applications A number of business applications require 2 Mbit/s service, which could be economically supplied by HDSL. This includes rapidly-growing market segments, such as videoconferencing, LANto-LAN interconnects and base stations for mobile communication. Another HDSL application is the provision of primary rate access ISDN. The interconnection of private PABXs via HDSL could also be a potentially worthwhile application. Clearly, HDSL is suited for use first and foremost in business networks, and there will be a very limited use for residential applications. However, an area that may allow the use of HDSL for POTS is the copper-fed street MUX (multiplexer). The upsurge in the demand for digital leased lines all over Europe offers a market for the connection of 2 Mbit/s lines to small and medium-sized businesses. A significant number of these applications will probably be implemented with HDSL, especially in areas where copper networks exist and where very fast ser-

156

vice provisioning at reasonable cost is required. For greenfield applications, fibre cables may be the first choice. In the short term, at least, the other major HDSL application will be for primary rate access ISDN. Main advantages of HDSL - fast provision of HDSL lines - prompt and easy delivery of 2 Mbit/s service to customers - cost-effectiveness - use of existing copper cable plant - no repeaters needed, which means improved data transmission quality - standardised performance monitoring of HDSL systems - reduced amount of electronics in the circuit (compared with PCM30 transmission technology or fibre technology) - HDSL transmission is less sensitive to cable characteristics

ADSL, Asymmetric digital subscriber line Today's media world is investing to build national information infrastructures - the "information superhighway". Telecom providers, cable distributors and a mix of other investors are developing their own visions of this highway. But consumers are still uncertain about how to find an onramp - the so-called "last mile" - that will connect them to access the large number of entertainment sources and other interactive capabilities envisaged. Most of the consumers are not specialised in data- or telecom and are therefore looking for a simple way to use the system. Several transmission methods are under discussion; some of them are currently in field trials. The three major methods are: Ericsson Review No. 4, 1995

- HFC = hybrid fibre coax (fibre fed to a distribution point and then coaxial cable down to the customer) - FTTC = fibre to the curb (fibre to a distribution point in the vicinity of the customer premises and copper for the last few hundred metres) - ADSL, using the existing copper infrastructure Each of these transmission methods has its own distinct strengths and weaknesses. It is likely that all of them will coexist and that the operator will benefit from the advantages of the various solutions. ADSL, for example, makes it possible to connect customers almost immediately using the installed copper cable plant. ADSL is also the most economical solution for the connection of a small number of customers to broadband services. Other possibilities are wireless systems and direct satellite connections.ADSL characteristics - transmission over unconditioned, twisted copper pairs - very low bit error rate (BER) - 1.5 or 2 Mbit/s transmission in subscriber (downstream) direction (6 Mbit/s or even more in the near future) - up to 64 kbit/s upstream (up to 640 kbit/s in the future) - independent POTS via PSTN - line code: DMT or CAP ADSL technology. ADSL technology - like other methods of transmission on subscriber lines - is the outcome of a decade of research and development in the field of digital subscriber lines. R&D activities have addressed the employment of adaptive signal processing for automatic adjustment of digital equalisers and echo cancellingfor precise compensation of noise, so as to allow the use of ADSL technology in the subscriber loop. The concept of ADSL is to provide higher bit rates exceeding 1.5 Mbit/s - towards the customer site. Several different transmission techniques and several different internal architectures have been proposed for asymmetric systems. One proposed transmission technique uses frequency division multiplexing to separate wideband downstream transmission from narrowband upstream transmission. This does not require echo cancellation, which may be advantageous Ericsson Review No. 4, 1995

from an implementation point of view. Since one direction of the narrowband channels is time division multiplexed with the wideband channel in the downstream direction, and since a forward error correcting (FEC) code will likely be applied to migrate impulse noise, this method has latency implications. Transmitting the narrowband channels in duplex mode, using echo cancellation for separate transmission of the wideband downstream at higher frequencies, is another solution. The strength of this method lies in its ability to decrease the latency on the narrowband channels. The most efficient transmission method and, hence, the one that offers the potentially highest performance, uses asymmetric echo cancellation - an advanced technique. These details and specifications of the modulation format are being addressed by the T1E1, which has published an interface standard for ADSL. At an early working group meeting, they agreed to define such a standard based on discrete multi-tone (DMT) technology. Besides DMT, two other technologies are used in practice: carrierless amplitude modulation (CAP) and quadrature amplitude modulation (QAM). The US standard has been defined as follows: - one unidirectional channel transmitting 6 Mbit/s over about 3.6 km on 0.5 mm copper or - one unidirectional channel transmitting 1.5 Mbit/s over about 5.5 km on 0.5 mm copper - up to nine 64 kbit/s duplex (a total of 576 kbit/s) or, alternatively, three duplex channels with capacities of 384, 160 and 64 kbit/s (a total of 608 kbit/s) - last but not least, the use of filters makes it possible to provide basic telephony service (POTS) as well. In Europe, no standard has yet been established, but the ETSI TM3 group responsible for local loop standardisation seems to accept the US standard (as they did with HDSL). However, the European standard will be based on the 2 Mbit/s structure. Another important feature of ADSL is that is has been designed to coexist with the present services in a telephone cable and, using a single twisted pair, provide normal POTS service in addition to full ADSL capability. ADSL systems will oper-

157

Fig. S Typical ADSL configuration with the ADSL exchange board installed next to the local exchange and the remote unit installed at the customer's. You can be engaged in a telephone conversation while watching a video on the T V .

ate even in a worst-case environment without any effect on services already put in place, which means that no special selection of cable is needed. DMT uses fast Fourier transform (FFT) to divide the transmission channel into subchannels. Because the phase response of each subchannel is linear, pulse distortion is minimised. The characteristics of each subchannel can also be measured, and the amount of data assigned to each one is varied according to its characteristics. This is particularly important on access lines, because there can be large digression between the transmission properties of different copper pairs in the same cable. The maximum bit rate possible is determined both by the transmission characteristics of copper and by any interference generated in neighbouring pairs. ADSL applications ADSL is the technology of choice for residential interactive broadband services (RIBS) to access the existing installed subscriber base, using standard localloop cables. A number of interactive broadband services can be offered - not only to residential users but to business users as well. Although a new technology, ADSL will certainly be capable of delivering the high bandwidth already mentioned. It is equally certain that there is a huge potential demand for the applications it could provide. However, several factors will ulti-

158

mately determine whether ADSL-based solutions will succeed: - pricing (of the service as well as of the ADSL devices needed) - competition from cable TV and video rental stores - competition from other technologies (such as fibre) - the regulatory environment The typical ADSL application is the videoon-demand (VoD) configuration, Fig. 5, consisting of the following components: - ADSL boxes - video server - set-top converter - user interface The success of cable TV has proved that there is agrowth market for television services, as a complement to the traditional broadcast channels. Both cable-TV companies and ADSL vendors (via PTOs) are keen to exploit the potentially massive market for VoD. Not only can VoD systems transmit videos to customers when requested; facilities such as pause, stop, rewind, etc. are also offered. But the system will not be restricted to ADSL on PSTN. Cable-TV companies are at least as eager to offer VoD, and for the user there will probably be no difference between the respective services provided. Other applications are: video games, teleshopping, tele-medicine, education, etc. Advantages of ADSL - ADSL lines can be provided very quickly Ericsson Review No. 4, 1995

- very fast and easy delivery of interactive broadband services - use of existing copper cable plant -very cost-effective A complete system consists of a central server, which has online access to selected material, and a tape-based storage library, which allows access to less frequently used material. Online (live) information sources can be processed centrally by online MPEG (moving picture experts group) encoders and fed into the network. The requested information can be distributed over an ATM or SDH network, which allows update of information server machines. The distributed information servers located in different central offices are responsible for subscriber authorisation, billing and provision of the information material requested (this material may be stored locally or downloaded from the central server).

VDSL In the spring of 1995, the T1E1.4 and ETSITM3 standard bodies agreed to start a study project with a view to developing a standard for VDSL. This was brought about by the fact that TOs continue to invest in deploying fibre to a distribution point near to the customer premises, but with the last 100 to 1,000 metres covered by copper cables. In such cases it will be necessary to have very high speeds over a short distance on the copper wires. There is still a question mark about the speed to be used, and the exact data rates have not yet been defined. In the end there may in fact be two versions of VDSL: - one with a speed of 25 to 26 Mbit/s

- one with a maximum speed of 52 Mbit/s The transmission distance should be between 100 and 1,000 metres.This distance is also related to the transmission speed. The questions regarding the line code to be used, transmission mode (asymmetric or duplex), etc. have to be raised later, when network and service requirements - and the noise impact- have been clarified.

Coexisting technologies in the access network Fibre, radio and copper do not exclude each other; they will coexist (and in some cases even need one another). There is room for all these technologies. Fibre deployment is still - and will continue to be - very slow in matching the installed base of copper pairs. Radio has always been a solution, and radio-in-the-local-loop systems are becoming increasingly important for narrowband applications.

Conclusion The access network of the future will not be uniform - it will be application-oriented and cost-oriented. The cost is a factor of vital importance to the TOs. They will always strive to use the most economical technology, and this technology will vary from application to application, whether fibre or radio or copper. Subscribers will not readily agree to pay for the cost of a new access network based on the newest technologies. It is therefore necessary to continue to use and further exploit the existing copper infrastructure.

BoxB Products PGS 0+2 PCM-2. PCM-4, PCM-8 0+10 / 0+12 HDSL Cobra HTU ADSL Cobra ATU 2 Mbit/s Cobra ATU !l 6 Mbit/s

Ericsson Review No. 4, 1995

MINI-LINK E - A New Link for Flexible Transmission in Cellular Networks Kent G Ahlqvist

MINI-LINK, the well-known family of compact microwave radios from Ericsson, is intended for point-to-point transmission of 2x2 to 1 7 x 2 M b i t / s ( 3 4 + 2 M b i t / s ) and is capable of covering transmission distances of up to 5 0 km per hop. The author describes the latest version - MINI-LINK E - which extends the range of applications as new frequency bands and higher traffic capacity are added to the concept.

Fig. 1 With MINI-LINK E, a wide variety of transmission requirements can be met. Here, MINI-LINK is used in both fixed and mobile networks. Private applications are also possible.

160

MINI-LINK microwave radios offer an efficient and fast way of developing transport networks for wide-area coverage in cellular telephony. The remarkable expansion in this field also means that the need for such radios is steadily increasing. MINI-LINK provides small- and mediumcapacity links for all kinds of applications, including fixed networks as well as temporary installations. Applications are found in both public and private telecommunications networks, as shown in Fig. 1. Temporary installations include those set up for sports events, exhibitions and conferences. MINI-LINK is used in some 85 countries all over the world

and the monthly production rate is more than one thousand terminals. MINI-LINK E adds a number of features to the MINI-LINK family - among them a radio for the 7 GHz band and 34 Mbit/s traffic capacity. The increased traffic capacity is available in all frequency bands, including 7 GHz and the previous 15, 23, 26 and 38 GHz. The high capacity in combination with flexible switching and traffic rerouting possibilities enables powerful network solutions. The control and supervision interfaces of MINI-LINK E are identical with those of the previous versions. This means that MINI-LINK E can be used together with MINI-LINK C in one com-

Ericsson Review No. 4, 1995

mon network, employing the same supervisory system. With the addition of the 7 GHz band, the hop length is increased, which means that MINI-LINK can be used in a backbone network and to achieve road coverage. The availability of different hop lengths and different frequency bands is shown in Box A. For a given project, the selection of frequency band is determined mainly by national regulatory conditions and requirements for hop length and availability.

System concept MINI-LINK E consists of a compact outdoor-mounted radio unit and indoormounted access modules. A wide variety of antennas can be selected to optimise each hop. Interconnection between the outdoor part and the indoor part is provided by a single coaxial cable, carrying the duplex traffic, DC supply voltage, service channels and operation and maintenance data. The principal components of a MINI-LINK E terminal are shown in Fig. 2. The main features of MINI-LINK E are: - A wide variety of configuration possibilities, ranging from unprotected (1+0) single terminals to nodes with several radio units in various unprotected (1+0) or protected (1+1) terminals, as well as repeater configurations with one common access module - Flexible and easy planning, installation set-up, commissioning and operation. This is achieved through integrated mounting of the radio and antenna with a single coaxial cable between the radio and the access module, in combination with software-controlled configuration and traffic routing - Excellent in-service performance with high system gain and spectrum-utilisation efficiency, thanks to a unique modulation process and advanced coding. The modulation process used in MINI-LINK E - constant envelope phase modulation, C-QPSK- is unique in that it combines the properties of constant envelope with high interference discrimination and is optimised for high frequency systems. To improve system gain further, forward error correction, FEC, has been incorporated as standard for all bit rates. Ericsson Review No. 4, 1995

Box A

Availability

Since microwaves are attenuated by rain, the dimensioning of high-frequency radio link connections must include a signal strength margin. This margin is determined by the following factors: -the frequency band used -the length of the connection -the climatic zone (expected rainfall intensity) -availability and quality requirements.

Fig. A The diagram shows the probability of the bit error ratio (BER) exceeding 10" 6 for MINI-LINK 7, I S , 2 3 , 26 and 38. The curves apply to areas where the rainfall intensity does not exceed 4 2 mm/hour for more than 0.01 % of the time.

For a given set of parameters (frequency, distance and climate) we can calculate the probability of the quality of the connection falling below a given level. Under normal operating conditions, a MINI-LINK connection is practically without bit errors.

Fig. 2 The main components of a M I N I - L I N K E terminal are the outdoor-mounted radio module and antenna unit and the indoor-mounted access module. Interconnection is provided by a single coaxial cable.

161

Fig. 3 A segment of a typical MINI-LINK E network where terminals with different traffic capacity are interconnected to form one common network and in this way optimise flexibility.

: -»£«'».;!». A r f | i ...jSAfSSK K£-

BOXB Data of the respective versions of MINI-LINK E MINI-LINK Frequency band (GHz)

7-E 7.1-7.7

15-E

23-E

14.4-15.4

21.6-23.6

26-E

Modulation

C-QPSK (constant envelope phase modulation)

38-E

24.5-26.5

37.0-39.5

RF output power standard, (dBm) high power, (dBm) Receiver threshold (10-3) 2x2 Mbit/s, (dBm)

+21 +28

+18

+20

+11

+15

+25

N/A

+18

N/A

-90

-88

-87

-87

-83

-85

-84

-84

-80

8x2 Mbit/s, (dBm)

-87 -84

-82

-81

-81

-77

34+2 Mbit/s, (dBm)

-81

-79

-78

-78

-74

4x2 Mbit/s, (dBm)

BOXC Data common to all versions of MINI-LINK E Traffic interface Supervision interface

CCITT Rec. G.703, balanced or unbalanced

Service channel interface

analogue 4-wire, 600 ohms 0.3-3.4 kHz

RS232C (V.24) serial interface digital 6 4 kbit/s, G.703

Performance monitoring

CCITT Rec. G.826

Power supply

41-72 V DC, option 24 V

Dimensions and weight Radio unit

411x326x129 mm

8kg

Antenna module 0.3 m

360x334x183 mm

4kg

Antenna module 0.6 m

660x660x402 mm

9kg

Access module 1U

44x483x285 mm

3kg

Access module 2U

87x483x285 mm

9kg

Access module 4U

176x483x285 mm

18kg

Environmental conditions

162

Outdoor units

Indoor units

Temperature

-40 to +55° C

-5 to +55° C

Relative humidity

8-100%

5-90%

Ericsson Review No. 4, 1995

The required bandwidth (channel spacing) is Channel spacing Traffic rate 3.5 MHz 2x2 Mbit/s 7 MHz 8 (4x2) Mbit/s 14 MHz 2x8 (8x2) Mbit/s 28 MHz 34+2 (17x2) Mbit/s The flexibility of MINI-LINK E makes it possible to realise almost any network topology, with very little demand for indoor space and a minimum of cabling - even in complex nodes. A limited variety of plug-in subunits can be combined in numerous ways to form nodes for different network complexity. Atypical MINI-LINK E network is shown in Fig. 3. Typical data of MINI-LINK E are presented in Box B and Box C. Radio unit The outdoor-mounted radio unit accommodates the microwave circuitry and is fully independent of traffic capacity. The unit is housed in a weatherproof and shielded box and has an IEC waveguide interface. This means that it can be mounted either to an integrated antenna, Fig. 4, or connected to any standard type of antenna. The box is painted light grey and equipped with a handle for lifting and hoisting, Fig. 5. The interface to the indoor unit is a single 50 ohm Type-N coaxial connector. For installation and service, the AGC voltage - corresponding to the input RF level - is available on a TNC connector. This signal is primarily used for antenna alignment. Two LEDs show the alarm status of the radio unit. The radio unit, which contains a microwave unit and a radio interface unit, is available for different frequency channel arrangements according to ITU-R and CEPT recommendations. The microwave unit comprises one microstrip board with an aluminium cover that provides shielded compartments for the high-frequency circuits. A control PCB is mounted at the back of the microstrip board. A DC/DC converter is connected via a flat cable. The microwave unit provides a waveguide interface to the antenna. The RFfrequency is generated by a processor-controlled synthesiser. Channel frequencies can be selected in steps of 0.25 MHz, either from a toggle switch on the indoor access module, or from a PC connected to the control and supervision Ericsson Review No. 4, 1995

Fig. 4 A MINI-LINK E terminal with a 0.3 m integrated antenna gives a minimum of windload to the mast.

interface. The wide subbands provided with MINI-LINK make frequency planning easy, since each unit can be set to a large number of different operating channels.

Fig. 5 The radio module is easy to handle as its weight is only 8 kg.

163

Fig. 6 The 0.6 m integrated antenna improves the hop length of MINI-LINK E.

The subband coverage is: MINI-LINK 7-E 80 MHz MINI-LINK 15-E 100 MHz MINI-LINK 23-E 550 MHz MINI-LINK 26-E 500 MHz MINI-LINK 38-E 280 MHz Antenna module The integrated antenna increases system performance, since no waveguide con-

nection is required. The radio unit can easily be dismounted from the antenna without affecting its alignment. The antenna modules for MINI-LINK E are available with diameters of 0.3 or 0.6 metres. They have an integrated radome and - like the radio unit - are made of cast aluminium, painted light grey. A 0.6 m antenna with radio unit is shown in Fig. 6.

Fig. 7 A 1U access module is used for termination of one hop.

164

Ericsson Review No. 4, 1995

High-performance antennas (with reduced side lobes) as well as standard antennas with and without radome are available. Radomes are used to protect the antenna from snow and ice. Antennas with other diameters, such as 1.2 m, are available as well as dualpolarised antennas. Antenna gain values for different sizes and frequency bands are specified in Box D. The antenna support for the compact antennas is made of aluminium profiles with stainless steel screws for fixing. The mounting support can be used for tubes with a diameter ranging from 50 mm up to a maximum of 120 mm. Elevation can be adjusted ±15°, and azimuth can be fine-adjusted ±40°. Access module The indoor access module can be dimensioned and equipped to suit different applications. As a minimum - for a 1+0 terminal station - only 1U (1U = 44 mm) is required in a 19-inch rack, Fig. 7. In a

Fig. 8 With a 4 U access module, up to four radio modules - serving four different directions can be connected together in a node. Traffic routing is software-controlled.

4U magazine, up to four radio modules can be connected, Fig. 8. The access module includes switching for protected systems as well as features for traffic rerouting at 2, 8 or 34 Mbit/s levels. Since the access modules are independent of the radio frequency band, one

BOXD Antenna gain MINI-LINK

7-E

26-E

38-E

N/A

1S-E N/A

23-E

0.3 m integrated, (dBi)

35

35

39

0.6 m integrated, (dBi)

N/A

37

40

41

45

1.2 m standard, (dBi)

37

42

46

N/A

N/A

Fig. 9 The flexible mounting kit makes it possible to mount M I N I - L I N K to almost any mast or tower structure. Ericsson Review No. 4, 1995

165

Fig. 10 In nodes with more than four radios, a number of access modules can be combined to provide the required functions.

module can be connected to radios operating on different frequency bands. Thus, compact node solutions are enabled. The access modules can be mounted in 19-inch ETSI or BYB racks. They can also be fitted into various types of Ericsson's radio base station, including RBS 2000. Fig. 10 shows one 4U and two

2U access modules, mounted in a 19inch rack. The access module accommodates three main types of indoor units: - Modem unit, MMU; providing traffic interfaces, signal processing and the interface to the radio unit. The MMU is available in different traffic capacity versions, for 2x2 Mbit/s,

Fig. 11 The MMU (modem unit) houses the baseband functions as well as the multiplexer for 2x2 or 4x2 Mbit/s.

166

Ericsson Review No. 4. 1995

4x2 Mbit/s/8 Mbit/s, 2x8 Mbit/s and 34 Mbit/s+2 Mbit/s respectively. The unit is equipped with a toggle switch and an LCD display for setting of the operating frequency and the RF output level. The display can also be used for local alarm and status monitoring. One MMU is shown in Fig. 1 1 . - Switch/multiplexer units, SMU; containing 8 / 2 Mbit/s and 3 4 / 8 Mbit/s multiplexer/demultiplexer, switches and control functions for 1+1 protected systems. - Service access units, SAU; providing parallel input/output ports, external alarm channel (EAC) interfaces and service channel interfaces. A SAU also contains an interface to the remote alarm channel (RAC), used to connect the operation and maintenance system for MINI-LINK clusters at different locations through modems or 64 kbit/s channels. The magazine provides mechanical housing and - through its backplane electrical interconnections between the indoor units. All external interfaces are located at the front of the units. The magazine is available in three different sizes to suit different node complexity.

Control and supervision Operation and maintenance of a MINI-LINK network is based on the powerful control and supervision system inherent in the MINI-LINK concept. Each terminal in a network is supervised by a built-in microprocessor, which communicates with the operation and maintenance (O&M) centre via a dedicated alarm channel, Fig. 12. This arrangement immediately notifies the O&M personnel of any alarm condition in the network. The following features are offered: - Alarm transfer channel - Performance monitoring - Near-end and far-end loopback test - Digital service channel(s) - Analogue service channel - Software-controlled traffic routing - Software-selectable output power MOMS/TMOS MOMS is a fully integrated application in TMOS and utilises the same workstations, databases, symbols and commands. It is used for centralised supervision of a complete network and provides network status presentation (alarm log, Ericsson Review No. 4, 1995

alarm list, alarm classification), line commands, check of authorisation, and network modelling utilities. MOMS supports all types of microwave radios belonging to the MINI-LINK family. The main objective of MOMS is to enable easy operation and maintenance of MINI-LINK networks by providing a userfriendly interface.

Fig. 12 Network management of a complex network with an operation and maintenance centre (OMC). Any terminal can be addressed and performance and status data can be read from the OMC.

MINI-LINK network manager The MINI-LINK network manager (MNM) is implemented in PC-based software that serves as a complement to MOMS for installation, configuration and fault location in the field. For smaller networks or for isolated clusters, the MNM can be used in an O&M centre. It offers facilities such as network status presentation, detailed alarm and status presentation, alarm log generation, command handling, automatic status polling and network audit. MINI-LINK E ring network In a MINI-LINK E ring network, data is normally transmitted in both directions. One of two possible paths is selected at the receiver side. The ring permits the two paths to be compensated for delays prior to switching, and, hence, makes for hitless switching. The ring network offers route diversity with fast synchronous switching at

167

Fig. 13 MINI-LINK is used to achieve good area coverage for cellular operators, in both urban and sub-urban areas.

2 Mbit/s level. One or more 2 Mbit/s data streams can be dropped and inserted at each site. Compared with conventional microwave radio networks, the ring concept provides higher availability to the user, at approximately the same cost. However, the frequency requirement forthe ring network is more stringent than for other topologies. This means that the cost of frequency licences must also be taken into account when a ring-network solution is considered.

and supervision system - offers new dimensions of flexibility in microwave networks. New frequency bands and high traffic capacity, within the same mechanical and functional concept, pave the way for efficient and flexible network planning and expansion. MINI-LINK E is designed to meet the most advanced requirements of network operators, today and in the future.

Summary With MINI-LINK E, Ericsson offers microwave radios with a wide range of frequency bands, traffic capacities, functions and features. The traffic routing function at the 2 Mbit/s level, integrated in the access module -together with a powerful control

168

References 1. Jungenfelt D.: New Generation of MINI-LINK Ericsson Review 70 (1993):4, pp. 132-139. 2. Lindstein T.: MOMS - An Operations System for MINI-LINK Ericsson Review 7 1 (1994):4, pp. 190-196.

Ericsson Review No. 4.1995

Dynamic Routing in Circuit-Switched Networks Hans Andersson and Mattias Lindeberg

More and more network providers have decided to abandon the hierarchical, fixed alternate-routing methods and are instead planning to introduce dynamic non-hierarchical routing for their transit networks. Real-time traffic control through dynamic routing has become a means of efficient handling of unplanned and forecast traffic variations. It is therefore worthwhile to investigate and compare the properties of the different dynamic-routing methods proposed by network providers. Methods such as Northern Telecom's dynamically controlled routing (DCR), AT&T's real-time network routing (RTNR), British Telecom's dynamic alternative routing (DAR) and NTT's state and time dependent routing (STR) are examples of proposed or already implemented dynamic-routing methods. The authors describe the above-mentioned routing methods and some other approaches towards dynamic routing. They focus on questions concerning network efficiency and resilience for the less complicated methods, compared with the centralised/distributed methods that use extensive network intelligence. The required processor capacity and the possibility of using analytical methods are also discussed.

Telecommunications have undergone sweeping changes in the past few years. Customers are offered new and more sophisticated services, business as well as social, thanks to the transition from analogue to digital telecommunication networks. This has increased customer demands and created new traffic situations; for example, excessive idle capacity in some parts of the network and overload in other parts. Increased signalling is also a result of the new services. Since these demands are becoming increasingly difficult to forecast, real-time traffic control (dynamic routing) is necessary to effectively handle unplanned and forecast traffic variations. A dynamic routing scheme provides flexibility at the switched level to adapt to changing traffic demands and shifts in

traffic patterns, and to reduce the effects of network failures. It is mainly used in non-hierarchical, inter-city, and international transit networks. The stored program control (SPC) type nodes in such networks are well suited for implementation of dynamic-routing methods. The tendency to build networks that are to a greater extent non-hierarchical also means that dynamic routing is of great interest to network providers. It is therefore worthwhile to investigate and compare the properties of the different dynamic-routing methods that exist today.

Dynamic-routing methods General overview The most commonly used method of rout-

Box A The following designations are used for the definition of the different dynamic-routing methods evaluated: N number of network nodes In number of circuits, link ij Xj: number of idle circuits, link i«->j Ky trunk reservation parameter, overflow calls

H M

maximum number of investigated two-link paths for an overflow call

Ericsson Review No. 4, 1995

Djj

alternate routing domain of link ie»j, i.e. the set of possible two-link paths i«->k«->j between the node pair \j or, for the sake of simplicity, the set of numbers ( k : l < k < N and k # ij) CRPjj currently recommended alternate path, used for the overflow traffic caused by the direct traffic i—>j Tjj threshold parameter for change of CRPjj, overflow calls i-»j

169

Fig. 1 Basic routing methods.

ing overflow traffic has been fixed routing. Every origin-destination (0-D) node pair has a set of predetermined alternative paths through the network. These paths are chosen once and for all, which gives poor flexibility for protection against different types of unplanned traffic variations. Dynamic routing is a means of finding the best way through the network, allowing the path of a call to vary, both with the state of the network and/or with time. The effect is more efficient use of the network, robustness and reduced cost for the network provider. A dynamic routing scheme should increase the rate of successful calls by increasing the number of paths the calls can take through the network. However, if the network is overloaded, the success of a call could block more than one other call, and so bring about a less satisfactory overall solution. There are two basic kinds of dynamicrouting methods: time-dependent and state-dependent routing, Fig. 1 . Time-dependent routing is based on the fact that traffic varies over the day: different parts of the network are busy for different periods. Different network configurations and different sets of possible alternate paths are therefore used for each origin-destination node pair during

specified periods of the day. An example of this is a network covering several time zones, with some parts of the network heavily loaded and other parts lightly loaded. Time-dependent routing is a means of handling forecast traffic fluctuations, but it gives limited protection against unplanned traffic variations if only fixed routing is applied within each scheduled routing scheme. State-dependent routing uses information about the present state of the network to decide which alternate path to use. A heavily loaded link should not be used for overflow traffic, since this may cause future blocking. The overflow traffic should be routed over the less loaded links. State-dependent routing adapts well to unplanned traffic changes. Three different methods can be used - isolated, distributed and centralised state-dependent routing - depending on how the information about the current state of the network is obtained and handled. When the isolated method is applied" each node in the network holds information about the states of its own links, Fig. 2. This information is used to decide which alternate path to use. When the distributed method is applied,7 each node holds the same information as in the isolated-method case and it can also interrogate the other nodes for the state of a specific link, Fig. 3. With the isolated/distributed methods, the intelligence is spread throughout the network. In order to determine the alternate two-link recommendation on a call-by-call basis, a distributed method may cause a significant extra load on the node processor, since this processor must handle extra communication and calculate and compare different path loads for each overflow call. When the centralised method is applied 89 , a central network processor collects information about the states of all links in the network every P seconds (P = 10 for example), Fig. 4. This requires a separate signalling network and a calculation centre, which makes the system more expensive and vulnerable to failures.

Evaluated methods Fig. 2 When an isolated routing method is used, the originating node holds information about the states of its own links only.

170

General assumptions In the following, a fully connected nonhierarchical mesh network with the nodes Ericsson Review No. 4,1995

connected by both-way circuits is considered. Direct calls are offered between each node pair. Overflow traffic is assumed to be routed according to a dynamic-routing method. If the direct routes between the nodes are termed links, then the set of possible alternate paths for each origin-destination node pair - the alternate routing domain - is equal to an arbitrarily ordered set containing all the possible two-link paths between the nodes in the node pair. Calls between nodes in a node pair are always first offered to the direct link between the nodes. If this link is blocked, the call overflows to a two-link path in the alternate routing domain. The rules for the selection of alternate paths in this domain make the difference between the dynamicrouting methods described below. Each dynamic-routing method uses some kind of trunk reservation, i.e. a trunk reservation threshold is applied to each link, and the directly routed traffic can use any idle circuit on a link, while the overflowing calls are accepted on a link only when the number of idle circuits exceeds this threshold. Trunk reservation prevents the overflowing calls on an alternate two-link path from blocking the subsequent direct calls on these links, since the overflow calls are only allowed to be carried by an alternate path if there is sufficiently high idle capacity on both links of the path. All the isolated methods evaluated here use a currently recommended alternate path, CRP, (i.e. the alternate path in the alternate routing domain which, at the moment, is the first to be selected and tested for an overflowing call). This path is retained as long as the calls overflowingto it are successful. Thus, the isolated methods are learning-automation methods which determine the new alternate paths to be used, based on the locally collected information on the success or failure of overflow calls. This type of methods perform best when some parts of the network are heavily loaded while other parts have considerable idle capacity. In a system with equally loaded links, the CRP should be changed randomly in the alternate routing domain in order to obtain the best network performance. Dynamic alternative routing (DAR) Dynamic alternate routing, DAR, is an isolated, dynamic-routing method developed Ericsson Review No. 4, 1995

Fig. 3 When a distributed routing method is used, the originating node holds information about the states of its own links and, by interrogating the other nodes, obtains information about the states of all links involved in the current connection.

by British Telecom for use in the United Kingdom inter-city network. DAR uses single-overflow trunk reservation on all links, and CRPs that are retained until the overflow calls are blocked. The trunk-reservation parameters are direction-independent, i.e. Kjj = Kj|. The DAR algorithm works as follows: - A call between an O-D pair is first offered to the direct link i k—>j) - If Xjk < Kjk and/or Xkj < Kkj, the call is lost, and a new CRPM is randomly selected from Djj. Otherwise, the call is successful and the CRPy is retained. Unsymmetrical dynamic routing (UDR) UDR is an isolated, dynamic-routing method. If the direct link is blocked, sequential multiple overflow is applied; that is, an alternate two-link path whose first link has not reached its blocking threshold is repeatedly searched for up to M times according to the currently recommended routing sequence (i.e. with sequential hunting for alternate paths in Dy, starting with the CRPy). Trunk reservation is only applied on the first link of the alternate two-link paths i—>k->j, keDjj, i.e. Kkj = 0 for all k e DB. Since a node - when an isolated method is used - holds information about the

Fig. 4 When a centralised method is used, a central network processor holds information about all states of all links in the network. This information and the current recommendations of alternate two-link paths are updated every 10 seconds, for example.

171

the call is offered the CRPjj ( i — ^ - H ) - If X ik < Kjkm for all M sequentially selected and investigated alternate paths i-»k m ->j, k m G Djj (1 < m < M), the call is blocked and the next alternate path in Djj is selected as CRPjj - If u. = min{m: 1 < m < M and x ik m > K i k J . H < M and X M = 0, the call is blocked and the alternate path i—»k„+1->j is selected as CRPy - If |i < M and X k ^ > 0, the call is accepted and the alternate path i—>k„->j is selected as CRPy.

Fig. 5 Network performance when different routing methods are used. 1 Direct traffic only 2DAR 3 UDR 4 UDR-CT 5 ALBA(2) 6 CLLPR 7LLPR Symmetric mesh network with N = 1 2 , L = 1 2 0 , and balanced traffic. M = 1 0 is used for methods 3 - 7 .

states of its own links but no information about the states of the "second" links in its alternate routing domains, and since crank-back (Box B) is not applied, the idea is to try to select a very lightly loaded first link (by using a large K), and to allow the overflowing call attempts to be successful as long as there is any idle capacity on the "second" link of the alternate twolink path considered. The UDR algorithm works as follows: - A call between an 0-D pair is first offered to the direct link ij - If XM = 0 (the direct link is blocked),

BoxB Crank-back is the term used for the function through which the originating node receives information from the via node about possible congestion on the "second" link, to be able to reroute the call.

172

Unsymmetrical dynamic routing with change threshold (UDR-CT) A drawback of DAR and UDR is the use of the CRPs until they are blocked, instead of changing them before they are blocked. Nippon Telegraph and Telephone (NTT) has therefore suggested the use of a CRP change threshold parameter5. This threshold parameter allows the algorithm to change CRP even if the overflow call is successful. UDR-CT is an isolated, dynamic-routing method using a CRP change threshold on the "second" links of its alternate routing paths; that is, if the overflow call is successful on the path i->k->j, and if Xkj < Kkj + T k j, then a new CRPy will be selected for the next overflowing call from the traffic i->j. Otherwise UDR-CT works exactly as UDR. Sequential multiple overflow and trunk reservation are applied for the first links of the two-link alternate paths, whereas no trunk reservation is used for the "second" link. If the "second" link is found to be fully occupied, the call is lost, since no crank-back is allowed. Aggregated least busy alternative routing (ALBA(2)) ALBA(2) is a distributed, dynamic-routing method which makes the alternate routing decisions on a call-by-call basis with only a limited knowledge of the states of the alternate two-link paths. Trunk reservation is applied to all links of the paths in the alternate routing domains. The trunk reservation parameters are direction-independent, i.e. Kjj = Kjj. See also Box D. Any two-link path i—>K—>j of M selected and investigated paths in the alternate routing domain is allowed to accept an overflow call if both of its links satisfy Xjk > Kjk and Xkj > Kk;. If this requirement is satisfied by more than one of the M Ericsson Review No. 4, 1995

investigated alternate paths, one of them is selected in a quasi-random fashion. If none satisfy the requirement, the call will be lost. If M < N -2, the M alternate paths are cyclically selected from the alternate routing domain (the currently recommended routing set in Dy with M elements is supposed to be stepped M steps (mod(N-2)) after each overflow call attempt from the traffic i->j). Least loaded path routing (LLPR) Least loaded path routing, LLPR, is a near-optimal, distributed dynamic-routing method that makes the alternate routing decisions on a call-by-call basis with an extensive knowledge of the states of the alternate two-link paths. Trunk reservation is applied to all links of the paths in the alternate routing domains. The trunk reservation parameters are directionindependent, i.e. Ky = Kjj. Any two-link path i—>K—>j of M selected and investigated paths in the alternate routing domain is allowed to accept an overflow call from the direct traffic if the states of its links satisfy the requirement: m/n[(X j k -K i k ),(X k j -K k j )] = = max{m/n[(X i k m -K i k m ),(X k m j -K k m j )]}, where max is evaluated out of the M elements i->k m ->j in the currently recommended routing set in Dy. It is assumed that the maximum is positive. If this requirement is satisfied by more than one of the M investigated alternate paths, one of them is selected in a quasi-random fashion. If the maximum is not positive, the call will be lost. Otherwise LLPR works in the same way as ALBA(2). Centralised least loaded path routing CLLPR is a centralised, dynamic-routing method that makes the alternate routing decisions on a nearly call-by-call basis with an extensive knowledge of the states of all the alternate two-link paths. Trunk reservation is applied to all links of the paths in the alternate routing domains. The trunk reservation parameters are direction-independent, i.e. Ky = Kjj. See also Box E. The alternate two-link paths are selected and investigated in a fashion similar to LLPR. The difference is that CLLPR determines the recommended alternate Ericsson Review No. 4, 1995

paths once every 10 seconds based on the information present at the update instants, and not call by call like LLPR. The assigned alternate path for each type of overflow call is kept constant between the update instants, and the overflow calls are successful as long as there are idle circuits on both links of the alternate path. This means that trunk reservation is not supposed to be applied in the nodes but only when the central network processor determines the routing recommendations. Otherwise CLLPR works in the same way as LLPR.

Fig. 6 Network performance when different routing methods are used. 1 Direct traffic only 2 DAR 3 UDR 4 UDR-CT 5 ALBA(2) 6 CLLPR 7 LLPR Symmetric mesh network with N = 12, L = 120, and 10% unbalanced traffic. M = 10 is used for methods 3-7.

Performance comparisions Compared with the isolated methods, the centralised/distributed methods perform best when the systems are lightly loaded and have a small number of circuits in the links. Asymptotically, all methods become equivalent as the number of circuits per link increases towards infinity. In Figs. 5 and 6, the six evaluated meth173

Fig. 7 Network performance as a function of the number of nodes N, all other parameters kept constant. Symmetric ALBA(2) networks with L = 1 2 0 , M • 1 0 , and balanced traffic of 1 0 8 . 6 eriangs per link.

BoxC E(L) e

lr

e

nc

Pe c lP c 2

mean number of circuits required per link mean number of link occupancies per direct call mean number of node through-connections per direct call mean processor power per direct call processor-dependent constants

ods are compared as functions of the link traffic A, when implemented in a rather small system with twelve nodes and links with 120 circuits each. In Fig. 5, all direct traffics are equal, while Fig. 6 shows an unbalanced traffic situation where the outgoing link traffic is equal to (A+0.1A)/2 for six nodes and equal to (A-0.1A)/2 for the other six nodes. In both

cases, the dynamic-routing methods are optimised for the direct traffic A = 108.6 eriangs. The evaluations are performed by simulations. Fig. 5 shows that the centralised/distributed methods perform better than the isolated methods when the links are lightly loaded. As the traffic increases, all methods asymptotically converge to the

Table 1 Network efficiency for the six evaluated methods of dynamic r outing N=12,

A=106.0.

Method

«•-)

«lc

No overflow

134.6

DAR

123.8

UDR

Q s =0.10 .,

N=12,

A=108.6.

Q s =1.00%

•V c 2