SE515054C2 - Improvements to or related to multicasting - Google Patents

Abstract

The present invention relates to multicasting in an ATM based public data network. Multicasting can be defined as the ability to send a single message to multiple recipients at different locations. A method of multicasting, of the present invention, operates in a packet switched data transmission system typically an ATM broadband data service system. A data packet, intended for transmission to a plurality of addresses, is copied at each node of the system to which one, or more, subscribers, to which said data packet is addressed, are connected. A copy of said data packet is transmitted to each addressed subscriber and to adjacent nodes to which addressed subscribers are connected. This procedure is continued until all addressed subscribers have received a copy of the data packet.

Description

30 35 551.5 0554 2 multicast/broadcast. Broadband data services (BDS) must also support multicast if the service is to be attractive to users. Group addresses are used to support multicast applications in these networks.

Typically, BDS is implemented at various ATM exchanges.

The "connectionless" server functions (CLSs) can be installed in, or in connection with, the ATM switches.

CLSs cannot be connected in all switching nodes. CLSs can form mesh networks, ring networks, star networks or tree networks. CLSs control the functions of the individual networks and forward data packets within networks and between networks.

Typically for ATM, an information packet, or cell, includes five bytes of header data and 48 bytes of user data. The header data includes data that identifies the cell, a logical address that identifies the routing, forward error correction bits, plus bits for priority management and network management functions.

Forward error correction is only used in the starting data.

All cells in a virtual circuit follow the same path through the network, which is determined during call setup. There are no fixed time slots, so each user can access the transmission medium whenever a free cell is available. ATM is capable of operating at bit rates of 155.52 and 622.08 Mbit/s.

Where a BDS, operating with a packet data transfer technology such as ATM, includes a plurality of switching nodes, the transmission of a plurality of identical data packets from node A, via nodes B, C, D and E to subscribers served from node E, involves a considerable waste of traffic capacity. The same information is sent between the same nodes many times. The present invention is intended to reduce this redundant transmission. 10 15 20 25 30 35 5155 054 The present invention uses in its implementation "agents" which are object-oriented programmed entities that include both data and operation code, and which have the ability to "negotiate" with other agents or entities, e.g. subscribers.

As is common practice in the telecommunications field, this specification makes extensive use of abbreviations. To assist the reader, the detailed description begins with a list of abbreviations used in the specification.

The present invention can of course be used with advantage in any packet-switched telecommunications network using mesh, ring, or tree network architecture in which the transmission paths are interconnected by switching nodes, some of which are CLS nodes, i.e. switching nodes having a CLS function.

However, the invention has in reality been implemented using an experimental BDS system developed by the applicant. For the sake of completeness, a brief description of the experimental BDS system is included in the present specification.

According to a first aspect of the present invention, a method of multi-addressing in a packet data transmission system having a plurality of switching nodes, each serving a large number of subscribers, is provided, characterized in that a data packet intended for transmission to a large number of addresses is copied at each node to which one or more subscribers to whom said data packets are addressed are connected, and a copy of said data packet is sent to each addressed subscriber and to the next node to which addressed subscribers are connected, until all addressed subscribers have received a copy of said data packet.

Preferably, a CLS node sends one and only one copy of said data packet to each CLS node: - to which it is directly connected; and - to which addressed subscribers are connected.

Alternatively, a CLS node sends one and only one copy of said data packet to each CSL node: - to which it is directly connected; and - to which either: - addressed subscribers; or - a CLS node connected to addressed subscribers; are connected.

Each multi-addressed data packet can include a group address.

A multi-addressed data packet transmitted from a first CLS node and intended for a large number of subscribers is transmitted from said first CLS node in a data packet encapsulating the group address.

Preferably, data packets are transmitted along tree-shaped paths; the packets are copied only where the branches of the tree divide, and in that the structure of said tree is determined so as to minimize system utilization, but with the restriction that the transmission delay to each group address is kept within a predetermined limit.

Said paths can form a Steiner tree.

According to a second aspect of the present invention, there is provided a data packet transmission system with a large number of switching nodes each serving a large number of subscribers, characterized in that each node includes a copying device for duplicating data packets intended for transmission to a large number of addresses, and a transmission device arranged to send copies of multi-addressed data packets to each addressed subscriber and to the next node to which addressed subscribers are connected.

Preferably, said transmission device is adapted to send one and only one copy of said data packet to each CLS node: - to which its CLS node is directly connected; and - to which addressed subscribers are connected.

Alternatively, said transmission device is adapted to send one and only one copy of said data packet to each CLS node: - to which its CLS node is directly connected; and - to which either: - addressed subscribers; or - a CLS node connected to addressed subscribers; is connected.

Each multi-address data packet may include a group address.

Said packet switching data transmission system may include at least one ATM-based public data network.

Preferably, at least one CLS node, in said packet switching data transmission system, includes group addressing means which for each group address has a database of all individual addresses that the group address comprises, said group addressing means being adapted to control a multicast addressing to a group address and to completely split a group address.

Alternatively, one and only one CLS node in said packet data transmission system includes a group addressing facility having for each group address a database of all individual addresses included in a group address, said group addressing facility being arranged to control multi-addressing to a group address and capable of completely splitting a group address.

Alternatively, one and only one CLS node, in said packet data transmission system, includes a first group addressing facility, said first group addressing facility being arranged to control a multi-addressing to a group address, and to partially divide a group address, in that each of a plurality of CLS nodes has a second group addressing facility arranged to complete division of a group address that has been partially divided by said first group addressing facility, in that only one first group addressing facility is allowed per group address, and in that more than one second group addressing facility is allowed per group address.

Embodiments of the invention will now be described by way of example, with reference to the accompanying figures, in which: Figure 1 illustrates the composition of an ATM address.

Figure 2 illustrates a centralized group address agent structure.

Figure 3 illustrates a path tree based on a source/destination path selection calculation.

Figure 4 illustrates the protocol architecture for a connectionless service that utilizes CL servers.

Figure 5 illustrates cell data processing in a CL server.

A glossary of abbreviations used in this specification is shown below: AAL3 - ATM adaptation layer, type 3 AAL4 - ATM adaptation layer, type 4 AAL5 - ATM adaptation layer, type 5 AC - access code ATM - asynchronous transmission mode B-ISDN - broadband integrated services digital network 10 15 20 25 30 35 515 054 2 BDS - broadband data service BOM - beginning of message CBDS - connectionless broadband data service CC - country code CL - connectionless CLNAP - connectionless network access protocol CLNAP-PDU - group addressed data packet, more generally a CLNAP protocol data unit CLNIP - connectionless network interface protocol CLS - connectionless server (connectionless server) CLSF - connectionless service function (connection- less service function) cnf - extension for confirmation extension for confirm messages COM - continuation of message CPE - customer premises equipment DG - designated gateway EOM - end of message FTP - file transfer protocol GA - group address GAA - group address agent GAP - group addressed data packet IP - internet protocol LAN - local area network LEC - local exchange code ME - mapping entity 10 15 20 25 30 35 s1ssøs4 q NDC - national destination code NGAA - nested group address agent NNI - network node interface OSPF - open shortest path first PDU - packet data unit PVC - permanent virtual circuit QoS - quality of service service) RD - routing domain req - extension for request message SMS - switched multi-megabit data service SN - subscriber number SSM - single segment message SVC - signaling virtual circuit UNI - user network interface VCI - virtual circuit identifier VP - virtual path VPI - virtual path identifier A multi-address information transfer enables a subscriber to send data packets to a packet data system which are transmitted by the system to multiple recipients. The present invention utilizes group addressing to achieve multi-addressing, and is described with respect to an ATM-based general data service. It will be apparent to those skilled in the art how to adapt the present invention to other packet-switched data transmission systems.

The system is required to deliver one and only one copy of a group-addressed data packet (GAP) over each of the connectionless access interfaces connected to the individual addressees represented by the group address. Each recipient of a GAP can use the destination group address, carried by that GAP, to multicast to the GAP's addressees (except its own).

Non-members of a group identified by a group address (GA) can send GAPs to that group. The service provider is responsible for assigning group addresses and ensuring that each GA uniquely identifies only one set of individual addressees. Within the network, the group-addressed data packets are structured into AAL5 flows.

The address structure used in the BDS in the present embodiment of the invention has the form: The address type C or E is a hexadecimal number that identifies the address as an individual address, or a group address. For example, C 4687075565H represents a BDS individual address, and E 4687075565H represents a BDS group address. The composition of ATM addresses is shown in Figure 1. The ATM address is first divided into address type and E.l64 number. The E.l64 number can then be divided into: - country code, CC, (1 to 3 digits); - national destination code, NDC; SN - routing domain, RD (1 to 2 digits); (3 to 4 digits); - subscriber number, (8 digits) - area 10 15 20 25 30 35 5 1. 5* Û 5-4 Ü - local exchange code, LEC (3 digits); and - access code, AC (5 digits) as shown in Figure 1.

A Group Address Agent (GAA) acts as a centralized administrator for a group address. The GAA interacts with a subscriber to handle requests for creation, modification, or deletion of group addresses, and provides information regarding the "membership" of an address. Group addresses can be stored in the BDS in three ways: - centralized storage; - replicated storage; and - encapsulated centralized storage.

In the centralized method of storing group addresses, one of the "connectionless" service nodes administers the assignment, deletion and modification of group addresses, and the inclusion, addition and deletion of individual addresses to the group, based on the user's instructions. From a numerical point of view, the GAA assigns a group a globally unique group address that belongs to the GAA's area. The GAA performs a complete address partitioning function. This approach has the following advantages and disadvantages: The advantages of a centralized approach are: - it creates a clear overview of an address group within a BDS system; - simpler administration and maintenance; - no need to implement special measures to ensure uniformity; and - it is easy to switch to a methodology with encapsulated group addressing. 10 15 20 25 30 35 515 054 fiâ The disadvantages of a centralized approach are: - delay time during transmission increases; - the load on an individual address agent increases; - the load on the network increases; and - lack of insensitivity to point failures (which can be improved by redundant agent implementation).

In the replicated, i.e. distributed, database method, the division of the group addresses requires that each network has its own database that must hold all the information regarding all the group addresses defined in all networks, regardless of their location. This method provides minimal latency during transmission because each copy is routed directly to its destination after the division of the group address in the originating network. It is insensitive because, if errors occur in the network, the remaining networks can divide the group-addressed packets.

In addition, fewer packets are generated. On the other hand, this method increases the network load because all multi-addressing occurs in the originating network. Maintaining consistency of "up to date" information is also difficult. For these reasons, this method is not part of the current description.

Encapsulated group addressing is used in conjunction with a centralized database structure. A group address agent manages each individual group address, as previously described for the centralized approach. Encapsulated group addressing can be used to avoid sending repeated information to individual members of a group that is accessed over the same network and uses an NGAA.

Since only one GAA is allowed per group address, one NGAA can be dedicated to serving each separate network that composes the BDS. There can be more than one NGAA per 10 15 20 25 30 35 5 15 /3 group address. Each network that has one or more members of the GA can have NGAA functionality. Thus, in the encapsulated group addressing method, the GAA performs a partial address resolution function. Encapsulated group addressing is an efficient technique for transporting group-addressed data packets. It results in lower network load because only one copy of the data needs to be sent to each of the participating networks.

However, it also causes increased transmission delay because the packets need to be routed to the central multicast network. Another problem with this method is a lack of insensitivity; single point failures can make it impossible to transmit or handle multicast packets.

The central database method is favored when the number of subscribers is relatively small. However, as the number of subscribers increases, the centralized method is upgraded to a system with encapsulated group addressing.

It is assumed that the ATM address in the GAA is well known to all subscribers of the BDS. Thus, a user can establish a control channel to the GAA to send requests for registration/deregistration of individual addresses and receive responses to the corresponding group addresses. The channel used to send multi-address packets (data channel) to the GAA may or may not be the same as the control channel. However, there may be some advantages in separating the control channel from the data channel, since the QoS and bandwidth requirements may be different for the different channels. Preferably, the control channel can be released after a predetermined time of inactivity has elapsed.

The protocol for registering a group address is as outlined below. 10 15 20 25 30 35 s1s 054 fff All individual addresses belonging to the same group are registered to a central GAA unit. A designated subscriber is an end user who is authorized to create a new group address. A designated subscriber is not authorized to create a new group address that excludes himself.

The following messages can be exchanged between a specified subscriber and the GAA device: Registersubs_reg (IndATMaddr1: IndATMaddr2: .....IndATMaddr3, transaction_ID) Registersubs_pnf (GroupATMaddress, transaction_ID, failure/success).

Non-group members can send packages to a group.

Confirmation messages are delivered only to the designated subscriber. (Alternatively, for security reasons, the confirmation message can be sent to all members of a group.) The Transaction_ID in the confirmation message should be the same as in the registration message. This ID is used for local purposes (e.g., validity of the confirmations). A given group address can have only one GAA.

The designated subscriber can, by interacting with the GAA unit, remove or add a new group member to an existing group of which he is a member. The designated subscriber can also ask the GAA to split a group address. This feature can be usefully utilized to check the current status of membership.

To prevent, among other things, non-group members from becoming group members without notifying other 10 15 20 25 30 35 s1seos4 [Y group members, only a designated subscriber is allowed to initiate the following command operation: Deletesubs_req (GroupATMaddress, IndATMaddr1: IndATMaddr2:........IndATMaddr3, transaction_ID) Deletesubs_çnf (GroupATMaddress, transaction_ID, failure/success) Addsubs_req (Group Aïnaddress, IndATMaddrl; IndATMaddr2:........IndATMaddr3, transaction_ID) Addsubs_cnf (GroupATMaddress, transaction_ID, failure/success) Resolvegroup_req (GroupATMaddress, transaction_ID) Resolvegroup_onf (IndATMaddr1: IndATMaddr2: ......IndATMaddr3; transaction_ID) A group address that has no members is automatically deleted by GAA.

Multi-address packets are sent out according to the following protocol: - source S sends a packet with destination address GA to the nearest connectionless node, e.g. N; - N identifies GA as a group address and encapsulates the original packet in a new envelope with source address (S,GA); - N forwards the packet to the central GAA unit; 10 15 20 25 30 35 515 os4i /é - GAA receives the packet and divides GA using the centrally managed address table; - GAA sends out as many packets as are needed for transmission to group members within the group; and - to transmit the unicast packets, the GA unit calculates the shortest path to all members of the group.

It should be noted that encapsulation means that the user data unit, when addressed to a remote node, is encapsulated as the payload of a new data unit whose header packet contains the appropriate addressing information it needs to reach its destination. The overlay of all the shortest paths calculated by the GAA unit forms a multi-address tree.

The detailed functions of this protocol can be more easily understood by considering an example illustrated in Figure 2. DG identifies members designated 1 to 9 in Figure 2. cLsa contains GAA for GA.

The operation of the centralized system protocol is as follows: - user S, connected to CLSa, generates a cLNAP-PDU with DA = GA(GA1>); - this PDU is routed to CLSc which contains the GAAn of GA; - CLSc divides GA into individual addresses for all members (i.e. 1 to 9): 10 15 20 25 30 35 515 054 f? - CLSc delivers the GAPn directly to its own members (5,6,7) and sends to each of the external members a CLNIP-PDU carrying the member's individual destination address; - no packet is sent back to the sender.

Multi-addressing, i.e. group-addressed packets are transmitted encapsulated in the data field of a CLNAP PDU. The GAA can transmit the individual packets either through point-to-point connections, or point-to-multipoint connections, where the GAA is the root of the tree and individual members are the leaves of the tree.

The multi-address packets are copied from the GAA to all other recipients, including network nodes, i.e. ATM nodes as well as connectionless servers. The multi-address packets are transmitted via paths that form a tree structure. The tree has its root at the GAA and the branches terminate at subscribers that are members of the group. The packets are copied when, and only when, branches of the tree split into two. Postponing the copying of the packets in this way conserves bandwidth in the network. If a single link leads to a multiple group of members, only a single copy of the packet traverses the link; if needed, it will be copied later.

Since the multicast path is computed as a tree at GAA, the path that a multicast packet takes depends on both the location of the GAA in relation to the source and its multicast destination. This feature is called "source/destination routing" and should be contrasted with OSPF (Open Shortest Path First) routing, in which the path is based solely on the destination. Taking the source into account when deciding on a path requires a lot of additional computation but leads to very good paths in terms of network utilization and delay for individual group members. In effect, the path taken between the packet source and each particular destination group member is the least costly available path. By using "source destination routing", as opposed to OSPF, a path similar to a minimal "spanning tree" is computed. This type of path is usually called a Steiner tree.

A typical path, corresponding to the centralized group address agent structure illustrated in Figure 2, is shown in Figure 3. The GAA, located in CSLc, constitutes the root of the propagation path. The branches terminate at group members 1 through 9. It should be noted that packets are copied when, and only when, two branches of the path go in different directions.

This results in a significant reduction in copying, which means there will be a significant reduction in data processing and bandwidth consumption.

Path calculations can be modified to take into account QoS required by particular applications or group members, such as delay, throughput, reliability, cost, operator, etc.

The present invention has actually been implemented on a prototype BDS based on the direct "connectionless" broadband data service model (CBDS) of the ITU/ETSI (1364) standard. This prototype will now be briefly described as a support for understanding how the present invention can be implemented. A connectionless service function (CLSF) is installed in the B-ISDN. The CLSF terminates the CL protocols and directs the CL packets to their destinations, according to control information in the CL-PDUs. The ATM connections between the user and the CLSF, as well as the connection between CLSFs, can be either PVC or SVC. The CL servers can be installed at different ATM exchanges, as well as at ATM cross-connect nodes.

The connectionless user sends data to a CL server whose address and identity it knows, e.g. the nearest one, to which it is either signaling or has a semi-permanent connection already established. The CL server then forwards the data to the target subscriber, possibly via a path that includes other CL servers. The CL servers can be interconnected via a virtual overlay network consisting of several VPs with pre-allocated bandwidth resources. The choice of PVC or SVC depends on the characteristics of the user traffic and the QoS requirements.

The use of CL servers in the ATM network will lead to a reduction in the number of required VPs (compared to a fully meshed VP network) and consequently to a concentration of connectionless traffic on fewer VPs. By statistical multiplexing of multiple sources on the same VP, burst transmission can be reduced although losses due to buffer overflows may increase. In addition, the number of connections required by each endpoint, for connection, is reduced to only one.

Figure 4 illustrates the protocol architecture on the CL servers at the user interface (UNI) as well as between the network nodes (NNI).

The CLSF terminates the B-ISDN CLNAP which includes functions for the structural description of the connectionless protocol to the connection-oriented ATM service using AAL3/4 units. The CLNAP includes functions such as routing, addressing, multicast, QoS selection, bearer selection, etc. The route selection is based on E.l64 address information contained in the CLNAP PDU header.

CLNIP supports the CL service between CL servers within a network operator area as well as between operator areas. CLNIP provides transport of both encapsulated and non-encapsulated data units. Functions for mutual interaction between CLNAP and CLNIP are provided by an ME in CL servers. 10 15 20 25 30 35 515 054 20 The CLNAP layer includes, among other things, functions for the routing and addressing of CBDS packets of variable lengths that are transmitted between a source and one or more destinations, without the establishment of any ATM connection by the user. It also supports multi-protocol encapsulation and the QoS parameters: transit delay, cost and residual error probability. To achieve higher traffic concentration at CLSs, VPs between CLs can be configured so that they do not have full mesh connectivity. Alternatively, the CL servers can be connected arbitrarily with each other, or using other topological systems such as hierarchical trees, buses or rings.

The service provided by BDS is quite similar to that of the IEEE MAC service: - BDS_UNIDATA.request (Source E.l64, Destination E.l64, User Data, QoS) - BDS_UNIDATA.indication (Source E.164, Destination E.164, User Data, QOS) This service has been implemented on a prototype BDS system consisting of a Fore System's ATM switch and 4 Sun/SPARC workstations. Each workstation is equipped with a Fore System's ATM data interface and a fiber connection to the switch. A workstation can be configured either as a terminal, or as a server. Implementation of BDS-related routines can be performed on the Fore System's host interface version 2.1.1., in a way that maximizes the use of existing Fore System's code for segmentation, reassembly, data structures, etc.

The terminal code consists essentially of an output routine and an input routine that construct and parse BDS 10 15 20 25 30 35 515 054 31 PDU start blocks, respectively. On the output side, the routine invokes Fore System's AAL5 protocol. Since this protocol requires a VPI/VCI pair to transmit on, a table lookup must be designed that provides such a pair corresponding to a given E.l64 destination address. On the input side, the AAL5 device delivers a buffer together with a BDS PDU start block that is parsed by the input routine and then forwarded to an appropriate higher-level device. Currently, two "upper layer" protocols have been taken into account: - IP for transparent use of current Internet applications such as FTP, virtual terminal (TELNET) etc.; and - a "Raw socket" interface that gives end users direct access to the BDS service described above.

Since the server operates on a "per cell" basis, the server code must be placed at the lowest possible level above the Fore System data interface. When a cell start block is retrieved from the ATM cell input FIFO queue, the server code checks whether the VP1 is one that is specified as part of the BDS network (VPI="CLS"), as shown in Figure 5. If so, a table lookup is performed to find the appropriate outgoing VPI/VCI pair. For efficiency reasons, the table mapping is maintained per session (using timeouts), rather than per packet, since it can be assumed that an application unit will generally generate more than one packet in its interactions with application units on other systems. If a cell is part of a COM, this involves an entry in a table.

More complicated procedures are required if a cell is a BOM or EOM, see Figure 5.

Claims (22)

10 15 20 25 30 35 40 515 054 '21 i PATENTKRAV A method of multi-addressing in a packet-switched data transmission system, having a plurality of switching nodes, each serving a plurality of subscribers, characterized in that a data packet, intended for transmission to a plurality of addresses, is copied at each node to which one or more subscribers, to which said data packets are addressed, are connected, and that a copy of said data packets is sent to each addressed subscriber and to the next node to which addressed subscribers are connected, until all addressed subscribers have received a copy of said data packets, wherein said packet switching data transmission system includes at least one ATM-based public data network, and in that one and only one CLS node in said packet switching data transmission system includes group addressing device which for each group address has a database of all individual addresses covered by a group address, where said group addressing device is arranged to control a fle address to a group address and to be able to completely divide a group address. Method according to claim 1, characterized in that a CLS node sends one and only one copy of said data packets to each CLS node: - to which it is directly connected; and - to which addressed subscribers are connected. Method according to claim 1, characterized in that a CLS node sends one and only one copy of said data packets to each CLS node: - to which it is directly connected; and - to which either: - addressed subscribers; or - a CLS node connected to addressed subscribers 10 15 20 25 30 35 40 51.5 054 23 is connected. Method according to one of the preceding claims, characterized in that each multi-addressed data packet includes a group address. Method according to any one of the preceding claims, characterized in that at least one CLS node in said data transmission system for packet switching includes group addressing device which for each group address has a database with all individual addresses covered by a group address, and in that this said group addressing device is arranged to direct a multiple addressing to a group address and to completely divide a group address. Method according to claim 4, characterized in that one and only one CLS node in said data transmission system for packet transmission includes a first group addressing device, wherein said first group addressing device is arranged to direct a multiple addressing to a group address, and to partially divide a group address; in that each of a plurality of CLS nodes has a second group addressing device arranged to complete the division of a group address which is partially divided by said first group addressing device; that only a first group addressing device is allowed per group address, and that more than one second group addressing device is allowed per group address. Method according to any one of claims 5 or 6, characterized in that a multi-address data packet originating from a subscriber connected to a first CLS node is sent to said CLS node, in a data packet encapsulating said group address. Method according to any one of claims 5-7, characterized in that a multi-address data packet is sent from a first CLS node and destined to a plurality of subscribers is sent from said first CLS node in a data packet encapsulating the group address. Method according to any one of claims 6-8, characterized in that said group addressing device, said first group addressing device, and said second group addressing device are agents, and in that said group addressing device and said first group addressing device are arranged to interact with a subscriber to handle requests for creation, modification or deletion of group addresses. 10 15 20 25 30 35 40 51 5 025.4 211 Method according to any one of the preceding claims, characterized in that data packets are transmitted along trees in tree form where packets are only copied where the branches of the tree divide, and in that the structure of said tree is determined so as to minimize system utilization, taking into account the restriction that the transmission delay for each group address is kept within a predetermined limit. Method according to claim 10, characterized in that said path forms a Steiner tree. A plurality of switching nodes, each serving a multi-packet data transmission system having a number of subscribers, characterized in that each node includes a copying device for duplicating data packets, intended to be sent to a plurality of addresses, and a transmitting device arranged to send copies of multi-addressed data packets to each addressed subscriber and to the next node to which addressed subscribers are connected, including at least one ATM-based public data network, and in that one and only one CLS node in said packet switching data transmission system includes device for group addressing, which for each group address has a database with all individual addresses comprising a group address, and in that said group address device is arranged to direct a multi-address transmission to a group address and to completely divide a group address. Data transmission system for packet transmission according to claim 12, characterized in that said transmission device is arranged to send one and only one copy of said data packets to each CLS node: - to which its CLS node is directly connected; and - to which addressed subscribers are connected. Data transmission system for packet transmission according to claim 12, characterized in that said transmission device is arranged to send one and only one copy of said data packets to each CLS node: - to which its CLS node is directly connected; and 105 20 25 30 35 40 sm 054 íS '- to which either: - addressed subscribers; or - a CLS node connected to addressed subscribers; is connected. Any one of claims 12-14, characterized in that each multi-data transmission system for packet transmission according to address data packets includes a group address. Any of claims 12-15, characterized in that at least the data transmission system for packet transmission according to a CLS node, in said data transmission system for packet transmission, includes a group addressing device, which for each group address has a database with all individual addresses comprising a group address , and in that said group address device is arranged to direct a multi-address transmission to a group address and to completely divide a group address. Packet switching data transmission system according to claim 12, characterized in that one and only one CLS node in said packet switching data transmission system includes a first group addressing device, said first group addressing device being arranged to direct a multi-address breathing to a group address, and to partially divide a group address; in that each of a plurality of CLS nodes has a second group addressing device arranged to complete division of a group address which has been partially divided by said first group addressing device; that only one first group addressing device is allowed per group address, and that more than one second group addressing device is allowed per group address. Packet switching data transmission system according to any one of claims 16 or 17, characterized in that a multi-address data packet emanating from a subscriber connected to a first CLS node is sent to said CLS node in a data packet encapsulating said group address. Any of claims 16-18, A packet switching data transmission system according to the characterization of a multi-address data packet transmitted from a first CLS node and destined for 515 054m SLL; to a plurality of subscribers is sent from said first CLS node in a data packet which encapsulates the group address. Data transmission system for packet transmission according to any one of claims 16-19, characterized in that said group addressing device, said first group addressing device, and said second group addressing device are agents, and in that said group addressing device and said first group addressing device are arranged to interact with a subscriber to handle requests for creation, modification or deletion of group addresses. Data transmission system for packet transmission according to any one of claims 12-20, characterized in that the data packets are transmitted via paths in tree form, where packets are only copied where the branches of the tree divide, and in that the structure of said tree is so determined that system utilization is minimized. that the transmission delay to each group address is kept within a predetermined limit. Data transmission system for packet transmission according to claim 21, characterized in that said paths form a Steiner tree.