AU698479B2 - Method for switching group addressed data blocks - Google Patents

Description

A

P/00/011 28/5/91 Regulation 3.2

AUSTRALIA

Patents Act 1990

OPIGINAL

COMPLETE SPECIFICATION STANDARD PATENT Invention Title: "METHOD FOR SWITCHING GROUP ADDRESSED DATA BLOCKS" 9 The following statement is a full description of this invention, including the best method of perfonning it known to us:o w e

II

L

L

This invention relates to switching data across a broadband network and and in particular to a method for distributing group-addressed data blocks in a network consisting of sub-networks as nodes and network-network connections as node connections of the network, with a cirti'al connection tree which contains all nodes to which the users of a defined group are connected, with the users connected to the same node forming a subgroup.

In broadband data networks the data to be switched are processed in a network node in accordance with preset protocols. The fragmented user data are provided with address information and service parameters. One such formatted protocol data unit (PDU) possesses, inter alia, a source address and a destination address. Each such data block is sent individually via a node connection of the network to one immediate node. On the basis of the carried address or service 1 5 parameter, what is to be done with the relevant PDU (protocol data unit) is decided in that'immediate node. In the most frequent case it is distributed to a further node connection or to a terminal unit interface; the latter takes care of the regeneration of the user data. Therefore there is no reserved connection between source and destination during data exchange. Chronologically staggered data blocks of different connections can run through the physically available node connections, this being described as a Connectionless Broadband Data Service.

It is possible for this mode of operation to operate with the data to be switched with suitable agreement about the addressing of a source from several practically simultaneous senders, the PDU being duplicated. An appropriately add* ssecl data block is therefore sent to a group of addressees, this being described as a group addressed protocol data unit (group addressed PDU, or GAP), as opposed to the individual PDU. A group address represents the set of all individual addresses which P are subscribers belonging, to the group. There are various possibilities for distributing such a GAP to all addressees of the group. One of them provides for routing each 30 GAP first of all to a central. node, where it is copied, each copy is provided with the individual address of the group menibr and is individually dispatched; the individual ~24j to-

C

I.

c c

E

f

E

C

O

rI

O

I

L

O

U

C

addresses of all group members also are held in the central node. Another provides for distributed storing of the group addresses in the sub-networks, which make up the broadband network. Each subnetwork knows the individual addresses of the separate group members and the adjacent sub-networks, in which further group members are located. An incoming GAP is distributed locally and to the adjacent sub-networks; the process is repeated in the latter. Each of the possibilities has its strong and weak points, which concern speed, system loading, failure tolerance, changeability and management of the groups and so on. A method with bundle group addressing is a combined form of the two possibilities mentioned and which shows a good compromise with regard to characteristics and is therefore used preferentially.

The broadband network consists of sub-networks, for example the MAN (Metropolitan Area Network), of the various network operators who for their part communicate with each other via two-way connections. The connections can also be linked via so-called "Connectionless Server" (CLS), turntables in transmission lines for passing on data blocks solely on the basis of their flags. For subsequent examinations,

CC

S such CLS are regarded as belonging to a subnetwork. Looked at as a design, the network consists of nodes (sub-networks, closed CLS) and node connections (two-way transmission lines). A subset of the node connections, through which all nodes of the 26 network can be reached, without developing a mesh, is known as a "Spanning Tree".

The node connections involved are branches in a tree which branch out into the nodes. An arbitrarily selected node forms the root then a hierarchical order follows from that.

QXCAP (Group Addressed Protocol data units), which reach a node from a source, S are passed on with bundle group addressing from there only towards the root node.

However on the way, the GAP are copied in the required number into the individual sub-networks, to which group addressees are connected, and are passed on to these addressees. The set of group addressees served by a subassembly is known as a bundle group. Such a bundle group is allocated a bundle group address. It identifies the subnetwork. The root node determines on the basis of th- available information which subscribers in the group still have not received a copy and undertakes distribution to these remaining subscribers.

G_

I1 The number of groups independent from each other, each with different combination with regard to the individual subscriber, is unlimited. The topographical location of the'subscriber can be widely different depending on the group, so that the definition of the tree is adapted with advantage. The "spanning tree" is therefore group-specific. But from that it can be assumed that it is defined for each specified group and is completely known at least in the respective root nodes, and in the remaining nodes as far as necessary. However, the information must still be considered more precisely, on the basis of which the root node has to decide which subscribers are still be served, yet it only knows directly from which node the group addressed data unit is coming, not however, on which path it reaches there. The GAP in fact contains an individual address, but states this in all of those cases of ambiguity over the path which the GAP has covered, in which a soL e, which is not a subscriber in the. group and therefore is unknown to the root node, has transmitted, which is specifically allowed. Furthermore, it should be noted that for internetwork connections, that is the data traffic over the node connections, the protocol provides encapsulation t cc SC of the PDU. For the processing of a data block arriving in a node, first of all only the PIL" information from there is available, which is in the capsule header of the encapsulated PDU. It should be possible to determine solely on the basis of this :16 information what ought to occur with the PDU.

The previously known method for determining subscribers, who have still not received a copy of the GAP which arrives at the root node, provides for the originator's address of the GAP in the capsule header to change once on its way to the root node.

A GAP initially always has an individual address which is recognised as such. One of 2S .the GAP, which for the first time reaches a subnetwork which produces the copies and distributes them to the addressees of the bundle group, and therefore a bundle group address is assigned to it, is added to the bundle group address as a sender address before transmission to the adjacent nodes towards the root node. An immediate subnetwork copies and distributes the GAP likewise as provided to the bundle group and to the nearest node towards the root node, but no longer changes the sender address. From the bundle address which refers to the subnetwork first dealt with, and with a knowledge of the tree, the root node determines which i i'i i; i i.

i ip I i 1 I. 1 l-^ f~'r 1 rf.

1 i

I'

i4 -4 subscribers in the group or which sub-networks have not yet been served with the GAP and orders the delivery.

In the known method the encapsulated GAP (GAPe) therefore contains in the encapsulated header as essential information the nearest bundle group address as destination address and the bundle group address of the subnetwork, which was the first to process the GAP, as sender address. It cannot be inferred from these addresses from which terminal unit interface the initial GAP originates; the information is contained in the sender address of the second cell in the header of the original nonencapsulated GAP (GAPn). The absence of this instruction in the first cell of the data block makes impossible an immediate decision on rejection of the data block, where applicable, if there is no entitlement to transmission. The opportunity for rejection is part of the service of "end-user blocking". It is part of the distribution process and should be given at each node connection. The distribution process should be able to take place solely on the basis of the information in the header cell of a data block ("on S the fly" process). There is a further disadvantageous aspect of the described process in "7.e that distribution occurs first in a branch only in a direction towards the root node, whereupon the GAP suitably addressed sometimes returns the same way.

The problem here consists in finding a method for switching group-addressed s data blocks which carries out the distribution on direct paths to the group members, without serving or missing out a member twice, and in addition it allows the decision o: to be made during processing of the first data cell of a data block whether it is to be passed on, processed, or rejected.

Distribution occurs gradually according to a repetitive instruction in accordance with the characterising features of Claim 1. The method uses selected network elements which are involved in distributing the data blocks in a known way as subgroup address agents. Local unique numbers, which are used as sender identification, are allocated to the subgroup address agents (SGAA) for implementing the distribution with data formats according to agreed protocols. The number of the dispatching SGAA is in each case written in the position in the header of the original data block which is not subject to any integrity checks. The significant information for the internetwork traffic in the first cell of a data block is not weighted by the

!I

distribution process, which makes possible the immediate decision about whether the arriving data block is to be passed on, processed or rejected.

In order that the invention may be readily carried into effect, embodiments thereof will now be described in relation to the accompanying drawings, in which: Figure 1: shows a structure of a data block with one GAPe; Figure 2: shows a diagram of a broadband network with a connection tree and node numbering.

Relaying of group-addressed data blocks takes place within the framework of the agreed protocols for the data exchange. A solution for efficient distribution of data blocks to a group must take account of the conditions as regards protocol.

Figure 1 shows the structure of a data block with one encapsulated groupaddressed protocol data unit GAPe. This contains a first cell 1 with 44 octets of 8 bits, a similarly structured second cell 2 and the user data 3. The first cell contains fouroctet long additional information 4 and the capsule header 5 of the encapsulated group-addressed protocol data unit 6, GAPe. Additional information 8 of the second e cell 2 joins on to it, after which comes the non-encapsulated group-addresses protocol data unit 7, GAPn, with its header 9 and the user data 3. The PDU is built up according to the protocol used for fragmenting the data and contains a destination 26 address field 12, DAn, a source address field 13, SAn, and service parameter fields 14. Destination and source address fields have the same format; they include the address type with four bits and the actual address with 60 bits. The structure of the latter is given by CCITT Recommendation E. 164, the generally valid numbering plan ''for the field of integrated services digital networks ISDN. The value in the subfield Address Type indicates whether it is an individual address or group address. The structure explained above is therefore also valid for individually-addressed PDU's, instead of the GAP the difference is only in the information contained in the address type field. The source address in the source address field 13 of the GAP is always an individual address and describes a particular interface, via which a user achieves access to the services. In this case a group'address is in the destination address field 12. In the capsule header 5 the destination address of the encapsulated PDU's DAe, and their source address, SAe, are available at the corresponding positions. The PDU's u

I

e;,tll .i:i i i :1 7 (equipped with the additional information 8) can operate unencapsulated inside a subnetwork, but encapsulation is not excluded. However, for transport between subnetworks, which have different owners, encapsulation of the PDU although not indispensable is necessarily provided from the protocol point of view. The destination address DAe is then either the address of the addressee in the subnetwork addressed or the group address. The addressee however can be a network element, which provides services for the network, in the case of interest here for example which carries out copying and distribution of data blocks with group addressing. In the cell section with the service parameter fields 14, two octet reserve 17 are available. These are not subject to any integrity checks and can therefore also be changed into encapsulated PDU's.

Further details of the data structure of the PDU are of no interest at this stage.

Group addressing (more precisely: the relaying of data units from a source to several users who are linked, as a defined group, to various sub-networks of a broadband network) is based on each user being open to conversation via a unique number, and elements of the sub-networks detecting functions which facilitate o ;distribution to a group: maintaining the group structure as well as duplicating and 4,t readdressing of PD'J's. However, account is to be taken of the requirements during implementation. On the one hand, the possibility of a data unit reaching the same addressee twice must be strictly avoided, while on the other hand the distribution should occur within a useful period, but without, improperly loading the network. Thus Sf all steps necessary for the distribution process should be feasible based on the first cell I ("on the fly processing"), but at the same time unauthorised users can be rejected at each access. The maximum allowable number of members of a group is limited just like the allowable number of groups.

Figure 2 shows the model of a network of the type described in the introduction, with sub-networks (including "connectionless server") in the nodes N1 to N13 and network-network connections as node connections C1 to C20. Network i 30 connections of subscribers in the form of user interfaces U1 to U3 are shown in '1 passing at the nodes N5 and N6. Countless connections of this type are available in the various sub-networks. Each interface globally has a unique number in accordance j i 1

C

P

D

C

C

r c o cc r

L

r with the numbering plan of CCITT Recommendation E.1 64. A particular selection among them defines a subscriber group. In the depicted example group members may be linked to the nodes N1 to N3, N5, N6, N8. N9 and N12. These nodes are highlighted here. In a continuation of the network (not depicted), which goes via the connection Cl 2 to the depicted network, at least one additional group subscriber may be connected. A more feasible connection tree for a group which includes all nodes with connections to these group members, is given by the bold marked node connections C1 -C6, C4-C1 2, C2-C7-C14, C1 5-C20 and C17. Alternative defined paths would have been possible, for example C14-C19 and C15 instead of C14 and C1 5-C20 or very radical C1 3 instead of C2-C7. The "leaves" of the tree are marked according to this defined path, i.e. the end nodes N1, N3, N5 and N8. First the root node is defined by free choice from the remaining nodes, in this case adopted by N12. The network element is located in the subnetwork of the root node, which perform special functions in connection with the distribution process during group C addressing; it is termed group address agent, GAA. The GAA undertakes in particular S the management of the group members and the connection tree. The connection tree C° I: is individually defined for each group for an immediate group it can look completely :different. One and the same node as a rule are part of several connection trees.

Since subscribers who are noi group members, also are permitted to send group-addressed protocol data units in the network, each node in the network, also N13 in the example in Figure 2, must be capable of encapsulating and sending at least one group-addressed PDU towards the root node of this group. In order to achieve efficient distribution, the nodes belonging to the connection tree must however '1 contribute more, while the PDU simply pass on to the root node. All group members connected to the one and the same node form a subgroup. The network element, which exercises a particular function in connection with group addressing, will in future be termed subgroup address agent, SGAA for short. The SGAA does not copy encapsulated group-addressed PDU's as frequently as" necessary. Further it encapsulates the GAPn, which contains the group address as destination address DAn, provides it at the same time with the destination address DAe in the adj-cent subnetwork and sends this GAPe there.

i A: r

A

r I-

S:-

i 9 An efficient distribution results if, in each node of the connection tree of the group in which a GAPn or a GAPe arrives, on the one hand the connected group members at most are served directly with the GAPn, on the other hand unlike the method with bundle group addressing all adjacent nodes of the connection tree receive a GAPe; the source, from which the GAP originates, is in this case always j removed from the distribution. Since the paths are clearly given in a distribution tree once established, the topological instruction described above ensures that no node or subscriber is served twice or even repeatedly. The root node with the GAA has no particular position in this, however it keeps one such position with regard to managing the group members and connection tree. As far as it is concerned, the instructions go to the SGAA, if changes are to be carried out. An additional agreement however can be reached for nodes which do not belong to the connection tree, such as N13 in Figure 2. A GAP, which originates from a subscriber not belonging to the group and 15 gets into the network via a node (N13) not belonging to the connection tree, must be passed on from this node only in a direction (via C9) with the destination address of i. the GAA in the connection tree.

For a distribution of this type in accordance with the additional abovementioned requirements regarding speed and accuracy, sufficient space for the "e2C necessary information is however unavailable in the ceil header. The following details are available to the nodes in the connection tree and have the given aim: Destination address of the SGAA, i.e. DAe; transport of the GAPe to the designated addressee, in this case the network element which is responsible for S* processing inside a subnetwork. Only the nearest SGAA can ever be addressed to observe the distribution instructions.

Source address of the SGAA, SAe; holding the first processed SGAA.

Source address of the user, i.e. SAn or SAe; source of the data. Facilitates, as SAe, the immediate rejection of data blocks which are unauthorised ("end-user blocking on the fly").

Group address, i.e. DAn; identification of a group. Individual address of the I;2 c.

GAA.

Conflict with regard to the source address arises iithe capsule header. In

JL-.

j~4)~~B

I,

i; i accordance with the invention, this is solved when the source address of the SGAA is replaced.

An analysis of the operations to be performed in implementing the desired type of distribution, in particular with regard to the chronological sequence of the steps, shows on the one hand the necessity of writing the source address of the user i.e. SAe in the capsule header, otherwise rejection of PDU's of unauthorised sources "on the fly" is not possible. On the other hand distribution of a SGAA is first of all carried out after it is established that it was correct as such and was justifiably addressed, for which the destination address of the SGAA must be in ihe capsule header in the internetwork traffic with encapsulated PDU's. Since one SGAA can serve several groups, the SGAA address states nothing uniquely about the group which is mentioned. That is why the information in the GAPn header must at least be taken into account for distribution.

Therefore it is sufficient to carry the information about the course of the data unit in the GAPn. Here however the source address SAn, which is unchangeable, is taken by that of the user who sends out the GAP.

o o As mentioned in the description of Figure 1, two octets reserve 17, which can o: be changed during encapsulation, are available in the header 9 of the GAPn 7. These are used there in accordance with the invention to identify the sending SGAA. Because 2E i distribution of a GAP is dependent on a stable network in which the relationships Sbetween the SGAA are absolutely fixed and known, it is sufficient to assign a local unique number to each SGAA. For this assignment there is no restriction whatsoever, except that it is made unambiguously and it does not go beyond the scope of the 16 bits available. Such a unique assignment has already implicitly been made in Figure 2.

'q2 In each subnetwork, which is depicted in the figure as a node which is a part of the connection tree, precisely one network element denoted as a SGAA is established. A local number, which corresponds to the numeral of the node marking in the figure, is allocated to this netw6ik element together with its global unique number which is r quired in eachcase as a destination address DAe. The local address of the SGAA in node N6 is therefore 6, that of the SGAA in node 9 is 9, etc. The 12 iTdicates the group address agent GAA. N4, N10 and N13 in fact fit into the local numbering scheme cnd the numbers 4, 10 and 13 could be used in this way as local SGAA iII Pij

C_

11 addresses for another group, but for the considered group could not part of the included SGAA.

The local number of the SGAA is in each case written in the two octets reserve 17 of the service parameter fields 14 in the header 9 of the GAPn 7 (Fig. The global number of the SGAA is not carried in the GAPn. Since the connection tree is obvious and each SGAA sends a GAP to all adjacent SGAA's in the group, except to the sender, the first processed SGAA does not need to be carried.

As an example which illustrates in detail in context the mode of operation, it is assumed that a user at node N5, but not that at the interface Ul, transmits a GAP. It is irrelevant whether or not he is part of the group. The transmitted GAPn contains the group address GA as DAn, and the individual address of the sending user as SAn.

As a destination address, the GA produces a switching of the GAPn to the subgroup address agents SGAA5 in the subnetwork N5. The latter recognises on the basis of instructions, which are available to it and which it has previously received from the GAA in its capacity as manager during configuring of the group, that it is a groupaddressed PDU which it must process in a certain way. First it identifies the user on the basis of the source address SAn and decides whether the user is to be blocked. If yes, it rejects the GAP, otherwy',9 local switching occurs to the members of the subgroup 20 and transmission occur. j the adjacent nodes, only to the node N6 in the example.

For local distribution, the SGAA5 copies the GAPn and sends it individually to S the user interface of members of the subgroup, to U1 in the example; it repeats this series,of events however frequently until all subgroup members are served, if the data i-t blocks came initially from a user who is a member of the subgroup, he does not receive a copy.

For transmission to the node 6, the SGAA5 writes its local unique number 5 into Sthe two octets reserve and encapsulates after that the GAPn, by adding the individual address of the SGAA6 as a destination address DAe and the copy of the source address SAn as a source address SAe in the capsule header.

The next processing of the GAPe in the SGAA6 of the node N6 includes the foilowing steps:

L.A

t' C ALCA TEL NV B^-f B.P. O'Connor (Authorized Signatory) P5/1/1703 8Ja.ne_1995 Date

W)

If) 0k j1 li:i i r, .I:i 12 1. The first cell is processed. The SGAA6 determines on the basis of DAe that it is mentioned, and on the basis of SAe, the user-source address, that no blocking .iist occur otherwise the processing would break off here. The first cell is no longer required further in the future.

2. The remaining GAPn is dealt with on the basis of the group address in the field of the destination address DAn and it is certain of the instructions available in the SGAA, that the users U2 and U3 make up the subgroup and that the surrounding nodes N2, N3, N5 and N7 are part of the group; on the basis of the local number 5, however, the SGAA5 in the node 5 is eliminated in the two octets reserve as an addressee.

3. The GAPn is copied twice and output to the user interfaces U2 and U3.

4. The local number in the two octets reserve is set on 6. The GAPn is copied three times and encapsulated, with the user address SAn as source address SAe and each individual address of the SGAA2 or SGAA3 or SGAA7 as destination address DAe. The resulting GAPe is output to the node connection C15 or C17 S: or C7.

In the adjacent nodes N2, N3 and N7, the four steps are repeated by analogy.

Nodes to be served no longer remain in the node N3 and the process ends there. In o i **6 1 the other nodes the process continues, until it stops in the end nodes everywhere.

I f a user (not belonging to the group) transmits a GAP to the node N13, then the group address produces an immediate capsule with copied destination address and source address as a destination address and the transmission to the node connection C9 the GAP is thus transmitted to the GAA in the node Ni 2. The same thing happens if a GAP is transmitted to the node N4. However since the SGAA6 can recognise the group address as such, it is possible to start with the distribution in N6.

A further improvement in distribution is achieved by the group address agent, GAA, addressing several nodes in one and the same branch on a direct route, by which various paths can be selected. Thus, for example, a GAPe arrived via the node N9 from GAA could be addressed to the node N3 via the node connection C2 to the node 7, on the other hand via the node connections C3-C9-C18, therefore via the nodes N10 and N13. This has the advantage of being able to select low-cost IT 1 ;II' 1.

1' SB.P. O'Connor A Ir8" I "i 1 1i 1 1 1 v 1 1 1 1 4 v 1 111 1 K 1 -1 1 1 1 1 1 1 .s l 13 connections, or being able to maintain the service during connection interruptions without the connection tree having to change. Obviously in the process the information must be propagated, which the distribution stops at the right place. Likewise, the two octets reserve can be used to this end.

A further advantage of the described solution to be mentioned is that it is not tied to the capsule of the PDU in the internetwork traffic. The GAP can be sent to the adjacent nodes as GAPn, provided that this is permitted by the protocol. The SGAA can make the first step described above also on the basis of the group address instead of the SGAA destination address the source addresses are identical anyway in the case of GAPe and GAPn. The additional steps are then independent in any case of the information in the capsule header.

0 4 iii 4 4 4 1 I'

Claims (7)

1. A method for switching group-adaressed data blocks in a network consisting of sub-networks as nodes and network-network connections as node connections of the network, with a virtual connection tree which contains all nodes to which the users of a defined group are connected, with the users connected to the same node forming a subgroup, wherein a distribution of group-addressed protocol data units is made by copies of group addressed protocol data units from the first node of the connection tree to all users of the subgroup, with the exception of the source, provided that it is a group member, and to all adjacent nodes, and furthermore in each node to all users of the corresponding subgroup and to all adjacent nodes, with the exception of the node from which the group-addressed protocol data unit arrived, wherein in each node of the connection tree a selected network element which is involved in the distribution of the group-addressed protocol data units acts as a subgroup address agent, each subgroup address agent being allocated a local unique number which is sent along to the next node as a sender identification of the group-addressed protocol data unit.

2. A method as claimed in Claim 1, wherein the local unique number in the SIt header of the protocol data unit is transported in two octets reserve which are not CI subjected to any integrity checks. 20

3. A method as claimed in Claim 1, wherein for implementation of the group, one of the subgroup address agents is designated as group address agent which manages all group members and the relevant connection tree and updates the subgroup address agent with the required information.

4. A method as claimed in Claim 3, wherein the group address agent is located in the root node of the connection tree. i

5. A method as claimed in Claim 3, wherein the group address agent sends group-addresspd protocol data units directly to several individually addressed subgroup address agents, with the group-addressed protocol data unit helping to transport additional information concerning tble distribution.

6. A method as claimed in Claim 1, wherein the group-addressed protocol data 'J i unit encapsulated for transport via the node connections carries the individual address i: i B B j ii i i: 1 1 8 I f~ i: i j F 17- 3 Ii- of the sending user as the source address in the capsule header, by means of which each node is capable of checking, on the basis of the processing of the first cell of the data block, whether any access authorisation is given.

7. A method substantially as herein described with reference to the accompanying drawings. DATED THIS FOURTEENTH DAY OF SEPTEMBER 1998 0 ALCATEL N.V. *e t 4 4 ec 4 4 t 4 C C CC 4 C.L C.. 44 t C C 1 such a GAP to all addressees of the group. One of them provides tor routing eac GAP first of all to a central node, where it is copied, each copy is provided with the individual address of the group member and is individually dispatched; the individual 1 1 1 ABSTRACT S"This invention relates to a method for switching group-addressed data blocks 4 in a network. In a broadband data network, consisting of sub-networks as nodes and network-network connections as node connections, it is possible to send protocol data units, termed PDU, practically simultaneously to several addressees, provided that these are defined as a group. Distribution of such group-addressed PDU, GAP, occurs with the assistance of subgroup address agents, (SGAA), in the individual sub- networks. All nodes which have user connections, which are part of the group, are integrated into a connection tree. The efficient and reliable distribution of a GAP takes place gradually via the connection tree as claimed in a repetitive instruction, which causes a copy of the GAP to be received in each node of the connection tree, in each Scase every one of the connections users or adjacent nodes being part of the group, S. with the exception of the sender. Lccal unique numbers, which are used as sender .identification, are allocated to the SGAA for implementation of the distribution with L~ data formats as claimed in an agreed protocol. The number of the sending SGAA is in each case written in the header of the original data block in the position which is not subjected to integrity checks. The significant information for the internetwork traffic in the first cell of a data block are not loaded by the distribution process, which makes possible the immediate decision about whether the arriving data block is to be passed *s on, processed or rejected. (Fig. 2) i o 4 I I ach cas wrtte inthehea er f t e rig naldat blck n t e p siton hic isno