CA2174402A1 - Method of operating a telecommunication network as well as network-access exchange and transit exchange - Google Patents

Abstract

The present invention concerns a method of operating a telecommunication network, which comprises a number of network-access exchanges and a number of transit exchanges, wherein each network-access exchange is connected directly, or via an arbitrary number of transit exchanges, to every other network-access exchange, and with a number of network-access channels wherein user data coming from the network-access channels of this network-access exchange are combined into data blocks in such a way, that each data block only contains user data intended for network-access channels of a single destination network-access exchange, that control data are assigned to the user data within the respective data block so that at the destination, the user data are assigned to a selected network-access channel, that a data block with address data is assigned to each data block with user data, and that each user data block within each exchange is switched transparently in the direction of the destination network-access exchange, based on the address data in the assigned address data block, and that a suitable network-access exchange and a suitable transit exchange are switched as well.

Description

~ ~` 217~02 Field of the Invention The invention concerns a method of operating a telecommunication network having a plurality of network-access exchanges and a plurality of transit exchanges, with each network-access exchange connected directly or via an arbitrary number of transit exchanges to every other network-access exchange and to a plurality of network-access channels.
The invention also concerns a network-access exchange and particularly a subscriber exchange, having a plurality of ports for network-access channels, particularly subscriber lines, and at least one port for a trunk to a transit exchange.
It is further directed to a transit exchange having a respective port for each trunk to another exchange.

Description of the Prior Art In the hierarchical telecommunication network of today, each call between a first and a second subscriber has its own tr~n.cmi~ion channel which is swilched via several exchanges or network nodes. This can be different in each section ofthe path as well, e.g. a synchronous time multiplex channel or also a virtual ATM
channel.
It must be possible to accurately access each individual transmission path bit by bit. This becomes more difficult as the transmission speed increases. In optical transmission technology, transmission speeds of more than 10 gigabit/s can be anticipated soon. To process such data streams, particularly to evaluate addresses and conve~t time positions, they are divided through demultiplexing into partial data streams, and these in turn are converted into slower data streams by series-parallel conversion, thus they are very strongly parallelized in the end. On the other hand, all prerequisites are provided to amplify the optical data streams, to switch l)el~eel- two paths or to convert between two frequencies or wavelengths.

~ ~7~402 Summary of the Invention The invention offers help in this respect by a method wherein at each network-access exchange, user data coming from the network-access ch~nnel~ of this netwolk-access exchange are combined into blocks such that each block only contains user data intended for network-access channels of a single destination network-access exchange, that the user data within the respective block are assigned control data to assign the user data to a selected network-access channel at the destination, that each block containing user data is assigned a block containing address data, and that within each exchange, each user-data block is switched transparently in the direction of the destination network-access exchange based on the address data of the associated address-data block.
It is also directed to a network-access exchange wherein means are provided for combining user data coming from the network-access channels into blocks such that each block contains only user data intended for network-access channels of a single further network-access exchange, that the user data in the les~eclive block are assigned control data to assign the user data to a selectednetwork-access channel at the destination, that the means are further adapted toassign a block containing address data to each block containing user data, that means are provided for transmitting user-data blocks and address-data blocks via a port for a trunk to a transit exchange, and that means are provided for receiving user-data blocks and address-data blocks from a port for a trunk and for outputting the user data contained in said blocks, based on the control data assigned thereto, via the port assigned to the selected network-access channel.
The invention is still further directed to a transit exchange wherein first means are provided for evaluating address-data blocks coming from a trunk and second means are provided for switching said address-data blocks, together with user-data blocks assigned thereto, to another port for a trunk.
The basic i~ea is to mesh all subscriber exchanges of the network by virtual paths. Each virtual path is implemented with a sequence of "supercollLainers", whose contents are tr~n~mitted transparently. Only the control information accol"~anying each of the "su~elcolltainers" is standardized and 217~0~

evaluated. In cases where this network does not reach to the subscriber, the network-access exchange is used instead of the subscriber exchange.
Finally, the subscriber exchanges are interconnected like the individual subscribers in the ATM (Asynchronous Transfer Mode). Data packets, called su~elcolltainers here, are formed, which have a packet header and an informationportion. The information portion is switched transparently; it can contain a number of ATM cells, for example, which come from subscribers in the output subscriber exchange area, and are intended for subscribers of the destin~tion subscriber exchange. The packet header carries the information required to guidethe respective ~upel`co~ iner to the destin~tion exchange. The tr~n~mi~sion speed in the paclcet header can be lower than in the information portion. This makes it possible to transmit the highest amounts of data on the one hand, and on the other to process the control information in the packet header on-line, with little effort in every exchange stage.
The packet headers need not necessarily be connected to the pertinent information portions in accordance with ATM cells. Other types of assignment are also possible. Insofar as the information portions are not stored temporarily, it must at least be ensured that the packet headers or the corresponding address-data blocks precede the information portions as user data blocks.
Several transmission paths, which run parallel through wavelength multiplexing, are used for the desired optical transmission. In this instance, it is possible to divide one of these tr~n.~rni.c~ion paths, i.e. a wavelength, into time slots, and to assign these to each of the rem~inin~ tr~n~mi~sion paths. The information portions are then tr~n~mitted on one wavelength each, while the packet headers are tr~n.~mitted in time multiplex on a different common wavelength. It can also be envisioned to combine a pair of broad band and narrow band transmission channels with separate wavelengths, and use them for information portions on the one hand, and for packet headers on the other. The use of separate tr~nemi.~sion paths for packet headers and information portions can facilitate evaluation of the addresses and other control data. However, care must then be talcen for the separate switching of these data as well, and for their clear-cut assignment.

217~02 A somewhat larger, albeit clear-cut gap b~ een packet header and information portion can be envisioned, for adapting the processing speed in the exchanges and the high tran.cmi~sion speed to each other. The gap can be filled for example by the information portion of a preceding packet, and the packet header of a following packet.
Configurating the exchanges according to the invention, and operating the network according to the invention, has the advantage that the serial user data stream can be retained, even at the highest tr~n~mi~ion speeds. Time position conversions within the network are not required, if each position can be switched to an alternative path. This provides the suitability for optical throughput.
Another signi~lcant advantage results from the fact that bit-by-bit synchronization is not required. All synchronization can be omitted in principle, if sufficient ~lt~rn~ive paths were always available. The lengths of the information portions could a1so be albitlaly.
However, a better utilization of the capacity results if fixed time slots are used. Yet, it is sufficient to insert a supercoll~iner into this time slot in a way so that a safety gap (guard time) is m~int~ined with the prece~1ing and the following supercontainer. The safety gap needs only to be large enough to prevent overlapping due to transit time fluctuations, which can no longer be equalized, and to enable safe switching belweell two supercontainers. The rem~inin~ portion of the time slot should of course be utilized as much as possible, so that the tr~nsmi~sion capacity can be used to the fullest extent.
Synchronization l"ea~n~s are therefore only required to the extent of ensuring this safety gap, and at the same time to be able to read and evaluate the addresses accurately. Such synchronization measures are simple to perform in theoptical field as well, for example with switchable delaying elements.
It is presently thought to con~lgure a supercontainer in such a way, that each of the user data contains a so-called VC-4 container according to the SDH
hierarchy.

~ ~ 2174402 A number of alternative tr~n~mission paths are preferably obtained by using a wavelength multiplex in each transmission path. But the multiplicity of the paths can also be used.

I~escli~lion of the Drawin~
The invention will be further explained in the following by means of a con~lguration example with the aid of the attached drawing, wherein:
Figure 1 illustrates a telecommunication network with network-access and transit exchanges according to the invention.
Figure 2 illustrates an optical switch fabric in an SDH/ATM environment, Figure 3 illustrates an opticat switch fabric in a WDM network environment, Figure 4 illustra~es a transparent optical WDM core network, Figure 5 illustrates s.lpelcolltainers in the optical core network, Figure 6 illustrates an optical WDM core network (optical packet switching techniques), Figure 7 illustrates contention resolution principles in the time and the wavelength domain, Figure 8 illustrates network and node architecture for the multidimensional switching concept, Figure 9 illustrates the basic switch module for wavelength and space switching, Figure 10 illustrates structure and tuning characteristics of the Y-laser, Figure 11 illustra~.es structure of the WDM switch module testbed, Figure 12 illustrates wavelength switching of containers over a wavelength span of 16 nm, Figure 13 illustrates space switching of containers and output signal form, Figure 14 illustrates supercontainers in an optical transport network, including the guard band between ~upel`containers~
Figure 15 illustrates the format of a supercontainer, Figure 16 illustrates an STM-16 output signal, Figure 17 illustrates a supercontainer output.

217~02 .
Detailed Description of the Preferred Embodiment The telecommunication network in figure 1 comprises network-access exchanges SF11, ..., ~FlN and STF1, ..., STFN, as well as transit exchanges SF21, ..., SF2N. The illustration in figure 1 only depicts one direction of tr~nsmission. Of course, each exchange also has the corresponding functions in the opposite direction.
Figure 1 depicts the entire telecommunication network as a single exchange. The illustration is also taken from an exchange description. This description was published as German Disclosure DE 42 09 790 Al, as European Patent Application 0 568 794 A2 and as U.S. Patent 5,369,514, all of which are hereby incorporated by reference. In actuality, according to the present invention, the entire telecommunication network is viewed as one large exchange, where the exchanges themselves are viewed like individual exchange stages. In principle, the internal construction of an exchange can be the same.
The entire network will not have the regular construction of an exchange.
Space position converters S, which can be realized as optical switches, are provided everywhere and are shown by the letter S associated with each network-access exchange and each transit exchange. Frequency position, or wavelength converters F, which can today be realized as optical components, are shown in figure 1 by the letter F and are provided with each network-access exchange and each transit exchange. Time position conve,lel~ T are only provided in the very last exchange stage, prior to leaving the network, and are shown by the letter Tfor the last stage of the network-access exchange. When entering and when leaving the network"he individual data streams must be adapted from the data format of the trunk area to that of the network, and vice versa. In any event, this requires temporary storage, time position conversions and changes in the transmission speed. As with the conventional exchange technology, known measures are provided to that effect at first.
The core part of future broadband communication networks will use optical high speed links with bit rates up to several tens of Gbit/s. In order to achieve a high flexibility in such networks, optical switch nodes can be used, with routing only optical signal paths rather than switching single connections. These 217~40 .
structures can be seen as the first step towards a transparent optical transportnetwork.
Switching in existing and actually planned broadband networks is based on time division multiplexing principles (PDH, SDH, ATM) both in the access and the core netwolk. This requires a high amount of buffering at any switch or multiplex stage throughout the whole network. Besides, by evolving toward Gbit/score transport networks to accommodate to growing needs, more and more problems with high speed signal processing on complex data structures will arisewith the need for tailored solutions. The introduction of optical switching in the core network using buffering and processing principles is severely handicapped by the fact that large scale integratable solutions, especially for optical memories, are not yet known.
This leads to wavelength division multiplexing ('NDM) core network architecl~r~s with transparent data paths avoiding multiple electrical/optical (E/O) conversions and repeated demultiplexing of Gbit/s data streams to access routingand control information. Instead, simple optical processing schemes and routing principles can be applied. Buffering in the time domain is thereby minimized or avoided.
The present invention discloses an optical WDM switching which provides optical paclket switching capabilities without buffers in the core of aswitch fabric. The wavelength domain is intensively used to resolve contention of packets by transferring the packets to another set of unoccupied wavelengths instead of delaying the packets in time buffer stages which are, as a consequence, only required at the borders of the network. A moderate extension in the number of wavelengths per link is required (e.g. to 5 wavelengths) to achieve the same llance as with 53 FIFO output queuing buffers. This principle is used on the core network level, where only large containers, each comprising a set of data units (ATM cells, SDH frames, etc.) in the payload and optically processable control fields, are handled in the wavelength domain without time buffering.
Buffers are only required at the borders to handle connections on the circuit layer.
Such a path routing `NDM network is of course capable of handling transparent WDM channels as well.

2174~02 The key building block for the described invention is a module for switching multiwavelength signals in the wavelength and space domain. A 4X4 system testbed for switching cells at 2.5 Gbits/s on 4 wavelengths or transparent multiwavelength channels, respectively, has been realized. Optical wavelength converter modules using Y-laser devices allowing wavelength setup times of 1.5 ns over a window of 20 nm and InP based multiwavelength space switch matrices with combination ani distribution capabilities, have demonstrated error free switching of cells separated by a small guard band of 3 ns. The same concept canbe applied to 16x16 modules for 10 Gbit/s signals.
The described system is a suitable basis for the beneficial introduction of high speed optical nt;lwolk~ exploiting WDM techniques instead of TDM with its extensive use of time buffer stages.
The Droblem of optical buffering and signal processing.
As stated above, switching in existing and planned broadband networks is based on time division multiplexing principles and protocols (PDH, SDH, ATM).
User information and related routing and control information is time multiplexedinto the same channel. This requires a high amount of buffering means for switching packetized or framed data. The control information has to be demultiplexed from the data stream in each switch module and very complex signal processing operations have to be performed after~vards. When the link speed approaches the multi-gigabit range, these operations become more and more problematic and their feasibility become questionable. It is not reasonable, however, to introduce optical switching in such nelwolks while ret~ining the time division multiplexing principles with their buffering and processing propellies.The required high thrbughput can be achieved, but the transparency gets lost dueto the fact that large scale integratable optical buffer memories of the required size are not known and complex optical signal processing operations are extremely dif~lcult to realize. As seen in Figure 2, to operate an optical switch fabric Awhich exploits the described capabilities of optical transparency and the wavelength domain in an environment of standard SDH/ATM interfaces requires a large effort in electronic circuits for SDH processing and demultiplexing, ATM
cell deline~tion and header processing, conversion into a suitable optical format ~ 17~4 .
and several E/O conversions. These circuits include sync demultiplexer SDH
outputs M[O, multiplexer and format converter MC, optical to electrical, format converter and electrical to optical circuit MT, MR, and associated input and output ports PA and PT respectively.
Network node interfacers for optical nodes.
The optimum environment to get the best possible benefit from the properties of optical switch fabrics leads to wavelength division multiplexing network architectures with transparent data paths. By this, multiple E/O
conversions and complex data plocessing can be avoided (see fig 3). Control and routing information can be separately accessible by optical means so that the Gbit/s user data stream remains untouched. By applying such a new optical transfer mode, the network node interfaces become quite simple and switching canbe performed preferably in the wavelength and space domain, whereby buffering in the time domain i3 nearly avoided. A transparent optical WDM core network can enable simple implementations of this network part, offering high speed optical data transfer without applying SDH/ATM schemes. The switching intelligence can be shifted towards the borders of the network into gateways andaccess switches providing the access to the network for standard SDH/ATM
signals, while leaving only reduced functionalities of routing WDM channels in the core. This core network with a flat hierarchical structure pe~ro.. lls the interconnection of the access nodes.
The bandwidth g~.
H[owever, there is apparently a gap between the bandwidth the user demands for provision of services, which could reach the range of about 150 Mbit/s per terminal ~in the access area, and the bandwidth of a transparent WDM
channel of up to several tens of Gbit/s, which is handled in the optical core network. Therefore it is reasonable to concentrate and multiplex the user traffic by ATM switches in the access area to fill WDM channels in the core network instead of assigning full WDM channels to single user connections which, as a consequence, make the network unfeasible. This helps to achieve better utilization of the network resources and to reduce the number of WDM channels to be handled by an access node. The task of the core nelwolk is limited to the routing 21 7~40~

of entire WDM channels between access nodes which could be ful~llled by a network structure applying WDM cross connect nodes (see fig 4). The network can be further partitioned to facilitate the network management and the routing operations.
The bandwidth ~ranularity of core network links.
The utilization of network ~esources, however, is still quite low in this kind of network architecture and the complexity of the interconnection scheme between the access nodes is high. This is caused by the fact that only full WDM
channels are routed and it is not possible to handle end-to-end connections in the core network. A multiplicity of connections with different destinations is multiplexed onto each WDM channel which cannot be split up in the network.
Consequently, each access node must simultaneously provide a high number of WDM channels, enough to interconnect with any other distinct access nodes to which a connection h~s to be established. This leads to a fully meshed or Clos-like interconnection scheme. Multi-hop connections via several access or gatewaynodes can reduce the interconnection complexity, but increase the required throughput per node.
The described approach for a simple optical WDM core network might be valuable for topologies with only very few nodes, where wasting some bandwidth in favor of a simple network implementation is acceptable. In arrangements with some thousand access nodes, pure WDM channel routing becomes inefficient and the approach of a transparent optical network should be extended towards introducing granularity and sharing the channel bandwidth among the data streamsheading for different destinations.
The most flexible and universal way to improve the granularity is the introduction of a novel optical transfer mode in the transport of the core network based on optical packet switching techniques. Thus the key characteristics of the WDM core network must be preserved.
Transparency.
Fully optical data paths without repeated E/O conversions are used. The packet oriented transfer mode can support a universal optical bearer service since the payload of the packets can be made service and bit rate transparent.

217~4~2 Exploitation of the wavelen~th domain.
WDM switching techniques have to be applied in the network nodes. No buffering is used in the optical core network. This implies novel solutions to the contention resolution problem.

Specific requirements regarding the processing of optical packets must be respected:
-Large packets (supercontainers, fig. 5) with optically processable routing and control information in the head are used. Short packets (e.g. ATM cells) at linkspeeds of several Gbit/s require too much processing power and extreme small switching intervals. The payload carries any kind of data transparently (SDI) frames, ATM cells, etc.). The supercontainers are separated by guard bands to reserve a set-up time for the switch elements and to allow for jitter and timingtolerances.

The ~u~ercolltainers have the preferred propellies as set forth in Table 1.
Container Based Transfer Mode.
Supercontail~ers in the optical transport nelwolk as shown in Figure 14 have the propellies set forth in Table 1.

- Long containers for reduced processing speed and effort.
- In core network only processing of entire containers.
- Access to payload only in LEXs at the borders of the core network (end of path).
- Header includes - optically processable SYNC pattern (for container and header synchronization) - address field with routing tag and control fields (for access to payload) - Payload is transparent with respect to protocols, data formats, bit rates, services (e.g. ATM cells, SDH frames, PDH frames, STM

~174102 signals, legacy services...) Bit rate of header and payload may be different - Containers separated by guard bands to allow for jitter and sync tolerances.

The format of the supercontainer is set forth in Figure 15.
Dimensionin~ of the SuFercontainers.
Dimensions of supercontainers has the following objectives:
- easy filling and access of the supercontainers - simple interfaces to other transfer modes - reduced control effort with large entities - high granularity of transport channel Dimensioning the Payload I,en~th.
The dimensioning of payload lengths provides:
Easy access of supercontainers which requires adaptation to the most used transport formats in other network partitions.
- SDH (include ATM in SDH) - Distributive services formats (MPEG, MPEG in SDH, analog) - Other, non-ATM formats SDH will be used on the majority of links as transport envelope.
Pure ATM to be used extremely rarely for transmission, since - ATM uses complex sync mechanism based on HEC (not feasible in Gbit/s range) - SDH is de~med as envelope for ATM and any mix of ATM with STM
- SDH aids network management functions and network supervision SDH is the major driving force for the length of the supercontainer's payload.
Basic Properties of SDH.
The basic properties of SDH are:
- Prima-~ data rate of SDH (STM-l) is 155 Mbit/s 217~0~
`
- STM-1 frame: 19440 bits, payload: 18720 bits (about 43 ATM
cells) - Repetition rate of the STM-1 frames: 9 kHz (relates to sampling frequency of POTS and frame rate of PDH) Basic question: How to treat the overhead fields?
- Translation to Supercontainer header fields - Mapping of fields into Supercontainer header The supercontainers have the preferred properties as set forth in Table 2.

Repetition rate of ~upelcolltainers of 8 kHz or multiples thereof.
Integer multiples of SDH channels mappable on a supercontainer path.
c The length of the payload must relate to one STM-1 frame (19440 bits) The bit rates of the payload and the header are chosen such that an integer number of SDH channels and the related sul~elcontainer header fields ~It into a tr~n~mi~sion frame.
Byte inlerleaved multiplexing. Some maintenance information of the section overhead is only transported in the first channel.
No access to STM-1 on SC level possible: Loss of granularity.
Each STM-1 frame is placed into one supercontainer.
The bit rate is adapted in order to fit integer multiples of STM-1 frames into 8 kHz frame.
Full granularity: Access to any STM-1 frame on SC level possible.
1) Transparent wavelength channels Routing of full wavelength channels (2.5...10...Gbit/s) Granularity not sufficient 2) Optical ATM
Direct transfer of formats and functions from electronic ATM to an optical implementation.

217~0 .
Use of ATM protocols for routing, signalling and OAM.

3) Supercontainers Optical implementation using simplified formats and functions.
Use of new optical bearer service adapted to technological and functional capabilities of optics.

2~ 74~02 !
The SDH multiplexing scheme from 16 STM-I to STM-16 is shown in Figure 16.
The multiplexing of STM-l signals to a supercontainer channel is shown in Figure 17.
A comparison of optical ATM and supercontainers is presented in Table 3.
T~BL~ 3 OPTICAL ATM SUPERCONTAINERS
Complex synchronization and cell Simplified sync mechanism with optical delineation procedure processing - HEC mechanism not feasible in - Guard band allows for jitter optics at Gbit/s speed tolerance - Bit synchronization required in - Only container sync required in each node transient core nodes - Cell processing with sub- - Relaxed processing speed nanoseconds precision - (several ns) High processing speed and effort Relaxed speed and effort - Very short cell dura~ion (43 ns - Large optical packets (e.g 125 at 10 Gbit/s) ,us) with guard bands - Switch reconfiguration within bit - Switch recon~lguration time 2 duration (< < 100ps) 20 ns Optical header processing without Optical header processing with alteration translation - End-to-end addressing - VC~/VP1 translation required No payload processing in transient core nodes Optical processing of ATM cell payload - Routing and OAM ~lelds only in (cf.ATM protocols) header - Signalling and OAM info Payload transparent w.r.t. bit rate, transported in payload format, protocols, services 2i744 .
Fixed bit rate for entire cell - Reduced header bit rate for easy opto-el. processing Very complex high order mux schemes Simple inte,~o,l~ g with existing - Complex interworking with standards present standards (ATM in - Fit complete STM-1 frame in SDH, STM-16 byte interleaved payload multiplexing etc.) No buffers in core network by using wavelength conversion - Transfer mode based on extensive use of buffers ~174~02 ~`
Other special requirements regarding processing of optical packets are:
-In the core network, entire containers are handled, the payload is only accessed in gateways/access nodes a the borders of the network.

-Simple synchronization schemes can be used to achieve supercontainer delineation. No synchronization on the bit level is applied in the core network.The related network structure is shown fig. 6. The WDM crossconnects are replaced by WDM supelcontainer switches. The interconnection links are dimensioned according to the actual required throughput between adjacent nodes instead of providing sufficient WDM channels to reach all destination nodes. This results in optimal utilization of the network resources and remarkable reduction of complexity (amount of nodes and interconnection links), but requires more sophisticated switch nodes.
Qptical WDM switching concept.
The switching concept to be applied in an optical WDM transport network should respect not only the basic characteristics of optical switching technologies like the transparency and the exploitation of the wavelength domain, but also the requirements imposed by the network architecture constraints as discussed above.The user requirement3 with respect to characteristics of the services they may demand is also an illlpollallt driving force. As a consequence, the switching principle has to support the handling of fully transparent wavelengths channels which may be offered to the user as well as the flexible switching of ATM/SDH
based data streams.
The multidimensional switching concept of the present invention therefore combines all dimensions of multiplexing; namely, in the space, wavelength and time domain. In general, each switching node can handle all switching domains.
In specific types of nodes, special and different emphasis can be put on the one or the other domain according to the network requirements and the actual status of optical switching technologies. In supercontainer switching parts of the network, for example, individ l&l supercontainers have to be switched quickly and contention resolution has to be performed. IN a path routing part, however, only semi-permanent switching of transparent wavelength channels is required. The multidimensional approach covers the entire transport part of the network from the access nodes in the subscriber access network to the nodes in the core.
In a supercontainer switching network, contention resolution is required in each switch module ~r any node and in each node of the network causing a high amount of buffer memories to achieve sufflcient low loss probabilities. Conceptsfor optical buffer memories of such large size (several tens of kilobits per module) and for Gbit/s signal in integratable and scalable form are not yet known. The present invention uses the flexibility of the multidimensional concept to avoid buffering in the core transport network. This is achieved by resolving the contention of supercontainers entering a switch element or node and competing for the same output in the wavelength domain. In contrast with time switching concepts (cf. electronic ATM switching) where the data units are delayed by a buffer memory and sent out one after the other, the new multidimensional conceptuses wavelength conversion to transfer containers to additional, unoccupied wavelengths so that ,he competing containers are sent out at the same time instant but on dif~erent wavelengths (see fig. 7).
This principle requires a certain expansion in number of wavelengths per link to be handled in the core network. An output buffer memory of about 50 queuing buffer places per link is applied in a time switching module to achieve the required loss ratio. This buffer can be replaced by an expansion to only 5 wavelengtlhs per link if a routing principle is applied which allows the routing to groups of 4 links. It is clear that such a loss system offers attractive opticalalternatives for queuing systems.
The descr;bed principle is applied to multistage switch nodes where all switch modules are interconnected with these multiwavelength links as well as tothe WDM core network where WDM links interconnect the switch nodes. The core switch modules and core switch nodes, respectively, do not embody any buffering. A buffer stage is only required in the very last stage of the transfer path through the network in order to concentrate again the previously expanded wavelength and to obtain the original signal format (see fig. 8). This stage is normally an access node to the optical core network where any kind of packet based data (ATM/SDH) is processed to put the data into the payload of ~ - 2174~02 supercontainers and to determine the destination access node address. Main partsof these operations are pelr~ led in electronics due to the complexity of the header processing and the sign~lling protocol. Consequently, the buffering operations are generally performed in electronics as long as optical buffering means are not available.
The basic functional block for the core switch nodes is an optical WDM
switch module with multiwavelength signals on input and output fibers whereby signals from any input fibre on any wavelength can be switched to any wavelengthon any output fibre. The module is capable of switching in the wavelength and space domain either in a fast operation mode to handle supercontainers where theswitch fabric has to be set up individually for each supercontainer within the small guard band between them, or in a semi-permanent operation to handle transparent wavelength channels which are to be established and released much less frequently on a call-by-call basis. Fig. 9 depicts a functional block diagram of such a switch module.

The main building blocks of the switch module are:
Address recognition and synchronization Olptical circuits are provided to extract the routing and control information from the optically processable header ~lelds of a ~upe~colltainer. The header information is just read and remains unchanged by passing through the switch module. Further circuits detect the start of a supercontainer, e.g. by optical correlation techniques and generate a container synchronization signal. Togetherwith a network wide master/slave synchronization scheme, a simple phase adaptionis sufficient for the synchronous processing of supercontainers, a bit level synchronization is not required.
Wavelength switching stage.
A key feature of the WDM switch module is the capability of pe~ ing all optical wavelength conversions. To obtain this feature, the WDM input signalfirst is split up in individual wavelength channels which are treated sepal~lely.
Optical signals on an arbitrary wavelength in the used wavelength window are transferred to a specific output wavelength which is determined by the module 2~7~40 .
control. This stage is optically transparent and is furthermore assumed to be used for the optical signal ~egeneration since the applied components encompass non-linear characteristics. The regeneration is essential for c~cading several modules and nodes and to construct a transparent optical network.
Space switching stage.
The space switch stage has the task of routing the individual wavelength channels to the destin~ion output ~Ibre which is prescribed by the module control.
To handle the multiwavelength output signals correctly, the space switch stage has combination and distribution capabilities.
Module conteol.
A control block implements the routing algorithm based on the routing tag attached to each supercontainer either individually per container for connectionless transfer or in a semi-permanent way on a call basis. Although parts of the control circuits - especially those located closely to the optical data path - may be implemented optically, electronics are used extensively in this block.
Only a principle module structure has been shown. Different topological arrangements of building blocks can be used depending on the requirements put forward by the applications (distribution capabilities, blocking plupe~lies, etc.).
Bxperimental results.
The following describes experimental results with key components for the described WDM switching concept in a system testbed showing the basic characteristics of switching optical contains at Gbit/s data rates.
Wavelength processing devices.
One of the key components is the tunable wavelength converter which allows the conversion of an incoming data stream from one wavelength channel to another. The performance of the wavelength converter covers the data transfer rate, the span of wavelength channels of the system and allows fast adjustment of the desire~ output wavelength. Different semiconductor optical amplifiers and DBR/DFB-based high speed wavelength converters operating at multigigabit data rates have been reported recently. The Y-laser device which is described here offers a large tuning r;3nge (~50 nm) and a fast wavelength switching speed (51 ns) at bit rates ~2.5 (3bit/s.

~t` 2174402 The Y laser is an interferometric laser type, realized with four independent segments and about 1.5 mm in cavity length. The vertical structure of the Y-laser contains several co-l-plessively strained quaternay quantum wells.
The lateral laser structure is fabricated with a buried heterostructure process. The electrical isolation between segments is obtained by etching the ternary contactlayer. Wavelength tuning of the interferometric laser can be achieved by ch~nging the effective length difference of the interfering branches. Varying the refractive index as a function of electrical carrier injection leads to the desired effect. The Y-laser is designed for precise channel adjustment in the range of 1540 to 1560 nm and for simple operation. This is achieved by introducing an asymmetric Y-laser type, allowing for single current tuning within the required wavelength range (see fig. 10). The fa3t tuning/switching between the de~lned wavelength channelscan be pel~o.llled in less than 1 ns.
Tunable optical wavelength conversion as the most important feature of the Y-laser device is described as follows: In a standard operation, the Y-laser is electrically preset to operate at )~out (arbitrarily selectable within the tuning range).
Injecting photons with )~in at an input port, we observe that output emission isswitched to the input wavelength as soon as the input power exceeds a certain threshold level (2100 flW). This effect is observed for input wavelengths being in the range of the laser gain spectrum. Thus, a large variety of input wavelengths leads to the described behavior. Filtering of the signal ~, at the output port maintains the signal ~,u.. which now contains the inverted bit pattern of )\,". In this way, tunable wavelength conversion over 45 nm is obtained, bit error rates of ~10-l2 have been achieved at bit rates 22.5 Gb/s. Signal regeneration capabilities can also be shown, extinction ratio regeneration by about 6 db is possible.
Finally, the t~mable wavelength converters can be available in a form that can be handled simply in systems. Y-laser devices have therefore been mounted in butterfly packages allowing for high speed operation. Using these modules, the relevant wavelength channels have been addressed in the system experiments, the ~Iber coupled power has reached values up to 0 dBm.
Space switching devices.

217~0 t`
The second key component is the optical space switch. The fundamental functions of splitting/combining and gating allow the set-up of all space switches.
The space switches presented here are consequently based on this principle. As afurther advantage, these semiconductor devices can additionally provide ampli~lcation. Taking advantage of such a modular concept comprising four elementary functions, it is feasible to realize space switches in a simple way and to extend these units froîn e.g. 2x2 to 4x4 in a compact form. Depending on the size of the switch matrix, complete monolithic integration or monolithic combined with hybrid integration (InP chips integrated on silicon motherboard) can be most appropliate.
The space switch architecture provides switching of each input signal to each output port (blocking free operation), distributive functions (one input signal to various output ports) and combining functions (several input signals to one output port). These functions are implemented for multiwavelength signals in a 20 nm range from 1540 to 1460 nm defined by systems application.
On the way to larger scale switch units, the basic functions are realized in monolithic integrated form: splittinglcombining, gating and amplification have been developed in lx2 switch elements similar to the Y-laser with all active elements, where segments 3 and 4 are used as gating/amplifying elements.
Antireflection coating can be used to suppress Fabry-Perot gain ripples of the integrated elements. ~xperimental results achieved so far with these elements show lossless operation (fiber-to-~lber insertion loss OdB), high speed (switching time <2 ns) and an extinction ratio of up to 30 dB. Furthermore, monolithic 1x4 switch elements have been realized with the aim to achieve blocking free 4x4 switch units by hybr-d integration. The required switching functions ON, OFF
and broadcasting ha~e been successfully demonstrated. For further c~cading of these switches, however, a configuration consisting of active/passive coupled waveguides appears to be app,~,p,iate.

System testbed results.
A system testbed was set up to check the feasibility of the WDM
switching concept based on the supercontainer transfer mode and to show the ~`
characteristics and pel~e~;live of present opto-electronic components technology in a system applications environment. The testbed demonstrates the key functions ofa WDM switch moduie; namely, the fast switching of optical packets in wavelength and space. With reference to practical signal formats, experiments have been carried out with short containers at a 2.5 Ggit/s and small guard bands between individual containers. This example can be considered as an extreme caseof shortest packets. A switch module capable of operating in such an environmentis well suited for any packet based application using larger packets like supercontainers or even semi-permanent paths.
The testbed structure is outlined in fig. 11. Programmable pattern generators and bit error measurement receivers serve as signal sources and sinks, respectively, for the 2.5 Gbit/s container streams. An electronic control block is provided to generate the high speed driving signals for the components and to control the sequence and timing of the experiments as well as the routing functions. Y-laser modules are used in the optical transmitter blocks showing the two main applications of these universal devices. In the first case, the Y-laser is used as a fast tunable wavelength source where the wavelength is controlled on acontainer-by-container basis. The second case encompasses a fully optical wavelength conversion of optical input signals to the Y-laser device. The outputwavelength is again controlled inclividually per container. The output wavelength of the asymmetrical Y-laser is determined by a single tuning current in a wavelength window of 1540...1560 nm. Due to the wide tuning range of the device, a channel spacing of 2 or 4 nm can be used leaving only moderate effort to the wavelength filtering and demultiplexing stages. Fig. 12 shows a multiwavelength outr,ut signal, where a single wavelength converter module is used to generate a data stream with 8 different wavelengths individually assigned to single containers over a span of 16 nm. The switching time of the digitally controlled tuning current is less than 2 ns, thus fitting in small guard band of 3 ns (equivalent to one byte of the 2.5 Gbit/s data stream).

For the multiwavelength space switch stage, only small modules of the size 2x2 or 4x4 in LiNb03 or InP technologies are available for the experiments. The main .~
217~0 .
functionalities as fast switching of signals in the used wavelength window and combination of these individual wavelength channels to obtain the multiwavelength output signals have been demonstrated in the testbed. Fig. 13 shows the spàce switching operation Or 2.5 Gbit/s containers. The switch set-up time is as fast as 520 ps leaving much resetve within the guard band. The eye diagram at the receiver input shows the excellent quality of the transferred data. Bit error free operation was achieved.
Today's standards for the broadband networks are characterized by SDH
and ATM. Both techniques need complex signal processing functions int he Line and Exchange Terminations as well as buffering inside the switch fabrics.
These functions cannot be implemented economically by means of optics, which is therefore an obstruction to the trend towards an optical core transportnetwork in the future. In order to enable the introduction of optical core networks anyhow, it has been shown that the problem can be solved by the application of optical WDM techniques in an optical switch fabric and by the definition of optical networks layers which can handle optically transparent channels. The signals arecarried on multiple optical wavelengths and are packaged in ~upercontainers consisting of e.g. multiples of ATM cells or SDH frames. This new optical transfer mode requires the introduction and standardization of novel network interfaces which are characterized by "simple" functionality in the core network, thus being suitable for optical implementation. Processing of SDH frames and ATM connections ate performed in the access nodes to the simplified core network only.
The basic technique applied in the core network is WDM. Concepts for routing signals in the wavelength and space domain have been described and experimental results on a key functional block, a wavelength and space switchingmodule using optical wavelength converts and optical space switches, have been presented to show the feasibility.

Other details can be found in the above inco-~ol~led by reference document.

Claims (8)

What is claimed is:

1. A method of operating a telecommunication network having a plurality of network-access exchanges and a plurality of transit exchanges, witheach network-access exchange connected directly or via an arbitrary number of transit exchanges to every other network-access exchange and to a plurality of network-access channels, characterized in that at each network-access exchange, user data coming from the network-access channels of this network-access exchange are combined into blocks such that each block only contains user data intended for network-access channels of a single destination network-access exchange, that the user data within the respective block are assigned control data to assign the user data to a selected network-access channel at the destination, that each block containing user data is assigned a block containing address data, andthat within each exchange, each user-data block is switched transparently in thedirection of the destination network-access exchange based on the address data of the associated address-data block.

2. A method as claimed in claim 1, characterized in that the user data are ATM cells to which control data have been assigned in the respective cell headers.

3. A method as claimed in claim 1, characterized in that the user-data blocks and the respective address-data blocks assigned thereto succeed one another in a data packet, with the address-data block used as a packet header.

4. A method as claimed in claim 1, characterized in that the user-data blocks and the address-data blocks assigned thereto are transmitted over separate, parallel transmission paths, and that the assignment follows from the relative temporal occurrence and the assignment of the transmission paths.

5. A method as claimed in claim 1, characterized in that synchronization measures are taken to cause the blocks to succeed one another without overlapping and without appreciable waste of transmission capacity and so as to read and evaluate the address data.

6. A method as claimed in claim 1, characterized in that for each transmission path, at least one alternative transmission path can be switched.

7. A network-access exchange (A), particularly a subscriber exchange, having a plurality of ports (PA) for network-access channels, particularly subscriber lines, and at least one port (PT) for a trunk to a transit exchange, characterized in that means (MC) are provided for combining user data coming from the network-access channels into blocks such that each block contains only user data (UD) intended for network-access channels of a single further network-access exchange, that the user data in the respective block are assigned controldata (CD) to assign the user data to a selected network-access channel at the destination, that the means are further adapted to assign a block containing address data (AD) to each block containing user data, that means (MT) are provided for transmitting user-data blocks and address-data blocks via a port for a trunk to a transit exchange, and that means (MR) are provided for receiving user-data blocks and address-data blocks from a port for a trunk and for outputting (MO) the userdata contained in said blocks, based on the control data assigned thereto, via the port assigned to the selected network-access channel.

8. A transit exchange (T) having a respective port (PT) for each trunk to another exchange, characterized in that first means (ME) are provided for evaluating address-data blocks coming from a trunk and that second means are provided for switching (MS) said address-data blocks, together with user-data blocks assigned thereto, to another port for a trunk.