Establishing connections across a communications network Number:7,394,811 from the United States Patent and Trademark Office (PTO) owispatent

Title: Establishing connections across a communications network

Abstract: On receipt of a request for a communication session over a communications network, such as an Internet Protocol communications network, a path for the session is established. In a preferred example the communications network is an MPLS network and the method uses a modified version of the SIP messaging protocol. Bandwidth along a chosen path is reserved and a messaging protocol such as CR-LDP used to establish this reserved path for the communication session. This creates a delay in the time taken to establish a communication session because the process of choosing a path and reserving bandwidth must be completed before the CR-LDP protocol is used. This delay is reduced by using modified SIP Protocol to perform the CR-LDP function too.

Patent Number: 7,394,811 Issued on 07/01/2008 to Gibson,   et al.

---

Inventors:	Gibson; Mark Robert (Bishop's Stortford, GB), Stacey; David (Starstead Abbotts, GB)
Assignee:	Nortel Networks Limited (St. Laurent, Quebec, CA)
Appl. No.:	10/825,541
Filed:	April 15, 2004

---

Related U.S. Patent Documents

---

	Application Number	Filing Date	Patent Number	Issue Date
	09410316	Oct., 1999

---

Current U.S. Class:	370/392 ; 370/400; 709/227
Current International Class:	H04L 12/28 (20060101)
Field of Search:	370/351,254,431,352,389,395,825,238,355,392,395.21,395.52,238.1,395.31,395.5,400,410,469,338,214-217,231,390,380,417 709/223,226,228,237,239,241,238,272,277,273,274,275,276 379/219

---

References Cited [Referenced By]

---

U.S. Patent Documents

5991301	November 1999	Christie
6075769	June 2000	Ghanwani et al.
6408001	June 2002	Chuah et al.
6735190	May 2004	Chuah et al.

Primary Examiner: Shah; Chirag G
Assistant Examiner: Mahmoudzadeh; Nima
Attorney, Agent or Firm: Barnes & Thornburg LLP

---

Parent Case Text

---

This application is a division of U.S. patent application Ser. No. 09/410,316, filed Oct. 1, 1999 now abandoned.

---

Claims

---

The invention claimed is:

1. A method of establishing a path in a communications network, the path being for use in a communications session between two endpoints, the method comprising the following steps: advertising information identifying a plurality of path elements or tunnels, the path elements or tunnels being between nodes of said communications network; at a first node associated with the first endpoint sending a communication session setup request message towards a second node associated with the second endpoint or with an intermediate node in the network; at a third node receiving the communication session setup message from the first node, or a message derived therefrom, and replicating the received message to form at least first and second forked communication session setup messages, the third node sending said first and second forked communication setup messages towards said second node along different respective paths; temporarily reserving bandwidth on path elements or tunnels traversed by the communication setup request messages; receiving at the second node at least two communication session setup request messages, these messages comprising, or being derived from, respective ones of said forked communication setup messages; selecting one of said different paths; at the second node sending a communication setup response message along said selected path towards said first node, said communication setup response message converting said temporarily reserved bandwidth on path elements of said selected path into permanent bandwidth reservations; and establishing a path for use in said communications session by storing items of said advertised information, the items identifying path elements or tunnels, associated with a path traversed by said communication setup response message; wherein the communication session setup request message sent by the first node towards the second node defines a partially explicit path between the first and second nodes, and wherein the third node sends the forked communication setup messages in response to the partially explicit path not defining a part of the path between the third node and the second node.

2. A method according to claim 1, wherein the third node determines that a node on the partially explicit path is not reachable through a fourth node and, in response, does not send a forked communication setup message to the fourth node.

3. A method according to claim 1, wherein the communications network is a label-switched communications network and the information identifying a plurality of path elements or tunnels comprises labels.

4. A method according to claim 1, wherein said information items are stored at nodes corresponding to endpoints of the path elements or tunnels associated with a path traversed by said communication setup response message thereby enabling data for the requested communication session to follow the selected path.

5. A method according to claim 1, wherein records of said respective different paths traversed by said at least two communication session setup request messages received at the second node are created as said messages, or messages from which they derive, traverse the respective different paths.

6. A method according to claim 1, wherein the first, second and third nodes are management nodes for transmitting control data and are each associated with respective abstract nodes for transmitting customer data.

7. A method according to claim 6, wherein the path elements or tunnels are between abstract nodes.

8. A method according to claim 7, wherein the abstract nodes comprise one or more physical nodes.

9. A method according to claim 1, wherein the communication setup request and response messages are based on the session initiation protocol.

10. A method according to claim 1, wherein said respective different paths are ranked according to their respective quality of service capabilities and the step of selecting one of said different paths is performed in dependence on said rankings.

11. A method according to claim 10, wherein said ranking is established on the basis of a combination of a first ranking determined by the first node and a second ranking determined by the second node.

12. A communications network comprising two endpoints and first, second and third nodes, the first node being associated with the first endpoint and the second node being associated with the second endpoint or with an intermediate node in the network, the network providing paths for use in a communications session established between the two endpoints, the network comprising: means for advertising information identifying a plurality of path elements or tunnels, the path elements or tunnels being between nodes of the network; the first node being arranged to send a communication session setup request message towards the second node; the third node being arranged to receive the communication session setup message from the first node, or a message derived therefrom, and to replicate the received message to form at least first and second forked communication session setup messages, the third node also being arranged to send said first and second forked communication setup messages towards said second node along different respective paths; the second node also being arranged to receive at least two communication session setup request messages, these messages comprising, or being derived from, respective ones of said forked communication setup messages; means for temporarily reserving bandwidth on path elements or tunnels traversed by the communication setup request messages; means for selecting one of said different paths; the second node also being arranged to send a communication setup response message along said selected path towards said first node, said communication setup response message adapted to convert said temporarily reserved bandwidth on said path elements into permanent bandwidth reservations; and means for establishing a path for use in said communications session by storing items of said advertised information, the items identifying path elements or tunnels associated with a path traversed by said communication setup response message; wherein the communication session setup request message sent by the first node towards the second node comprises a data structure defining a partially explicit path between the first and second nodes, and wherein the third node is arranged to send the forked communication setup messages in response to a determination that the partially explicit path does not define a part of the path between the third node and the second node.

13. A communications network according to claim 12, wherein the third node is arranged to determine that a node on the partially explicit path is not reachable through a fourth node and, in response, to not send a communication setup message to the fourth node.

14. A communications network according to claim 12, wherein the communications network is a label-switched communications network and wherein the information identifying a plurality of path elements or tunnels comprises labels.

15. A communications network according to claim 12, wherein nodes corresponding to endpoints of the path elements or tunnels associated with a path traversed by said communication setup response message are arranged to store said information items thereby enabling data for the requested communication session to follow the selected path.

16. A communications network according to claim 12, wherein it includes means to create records of said respective different paths traversed by said at least two communication session setup request messages as said messages, or messages from which they derive, traverse the respective different paths.

17. A communications network according to claim 12, wherein the first, second and third nodes are management nodes for transmitting control data and are each associated with respective abstract nodes for transmitting customer data.

18. A communications network according to claim 17, wherein the path elements or tunnels are arranged between abstract nodes.

19. A communications network according to claim 17, wherein the abstract nodes comprise one or more physical nodes.

20. A communications network according to claim 12, wherein the communication setup request and response messages comprise data structures based on the session initiation protocol.

21. A communications network according to claim 12, wherein means are provided for ranking said respective different paths according to their respective quality of service capabilities and the means for selecting one of said different paths is arranged to perform the path selection in dependence on said rankings.

22. A communications network according to claim 21, wherein said ranking means establishes rankings on the basis of a combination of a first ranking determined by the first node and a second ranking determined by the second node.

23. A computer program stored on a computer readable medium, the computer program being for controlling a communications network comprising two endpoints and first, second and third nodes, the first node being associated with the first endpoint and the second node being associated with the second endpoint or with an intermediate node in the network, the computer program executed by a computer controlling the network to establish a path for use in a communications session between the two endpoints by performing the steps of: advertising information identifying a plurality of path elements or tunnels, the path elements or tunnels being between nodes of said communications network; at a first node associated with the first endpoint sending a communication session setup request message towards a second node associated with the second endpoint or with an intermediate node in the network; at a third node receiving the communication session setup message from the first node, or a message derived therefrom, and replicating the received message to form at least first and second forked communication session setup messages, the third node sending said first and second forked communication setup messages towards said second node along different respective paths; temporarily reserving bandwidth on path elements or tunnels traversed by the communication setup request messages; receiving at the second node at least two communication session setup request messages, these messages comprising, or being derived from, respective ones of said forked communication setup messages; selecting one of said different paths; at the second node sending a communication setup response message along said selected path towards said first node, said communication setup response message converting said temporarily reserved bandwidth on path elements of said selected path into permanent bandwidth reservations; and establishing a path for use in said communications session by storing items of said advertised information, the items identifying path elements or tunnels, associated with a path traversed by said communication setup response message; wherein the communication session setup request message sent by the first node towards the second node defines a partially explicit path between the first and second nodes, and wherein the third node sends the forked communication setup messages in response to the partially explicit path not defining a part of the path between the third node and the second node.

---

Description

---

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to a method of establishing a connection between two endpoints in a communications network, and in particular but not limited to, situations where it is required to reduce the time taken to establish the connection. The invention also relates to a communications network within which this method is implemented and also to a computer program for controlling a communications network in order to implement the method.

2. Description of the Prior Art

One factor that must be taken into account during design of communications systems is that of the time taken to establish a communication session. This should be as short as possible in order that customers are not deterred from using the communications service and for applications in which communication sessions must be established without delay.

The time taken to establish a communication session is a particular issue when providing a guaranteed quality of service for transmission of internet protocol traffic. Quality of service is an important factor; customers require a good quality of service for message transmission especially for real-time applications such as video conferencing and voice. As well as this many customers require a particular level of quality of service to be guaranteed; if quality of service drops below a certain level and transmission is interrupted or noisy this may be acceptable in some situations but unacceptable in others. If particular levels of quality of service can be guaranteed this is particularly advantageous. A number of approaches to provision of guaranteed quality of service for transmission of internet protocol traffic are now described and the time taken to establish such connections discussed.

One approach that has been used is to prioritise individual transmissions that are sent over the network. For example, a system known as "DiffServ" allows messages to be marked to indicate their priority. Nodes in a communications network are then arranged to process high priority messages first. This enables high priority messages to be processed quickly but it does not provide a guaranteed level of quality of service. Also a certain amount of processing time is necessarily taken up in determining the highest priority message at a given node and this can lead to increases in the time taken to set up a communications session.

Another approach has been to reserve bandwidth over a particular route in a communications network. However, systems that use this approach (for example RSVP Resource reSerVation Protocol) typically are poor at implementing aggregation mechanisms--for example they cannot easily combine a number of separate sessions over the same route, each must have its own reservation. Another shortcoming is that they also typically only allow the called party to reserve bandwidth that is required to host a communication session. This does not allow the calling party to specify their requirements and this is problematic, especially because the calling party is typically the party which incurs costs for a call. Also, the time taken to set up reservations is often significant and can introduce delays in the time taken to establish a communications session.

Multi Protocol Label Switching (MPLS) is a standard messaging protocol that is suitable for carrying Internet Protocol traffic over communications networks such as Asynchronous Transfer Mode (ATM) networks and Frame Relay networks.

Constraint-based Routing Label Distribution Protocol (CR-LDP) is also a standard messaging protocol (CR-LDP is defined in Internet Draft: draft-ietf-mpls-cr-ldp-01.txt) that is suitable for use with communications networks that use MPLS. Mechanisms such as CR-LDP allow MPLS the ability to set-up paths between two endpoints over a list of routers, where these paths have ATM-like traffic requirements. However, there is no well-defined mechanism for the choice of the routers in this path that makes full use of the ATM-like traffic parameters. The only existing mechanism (QOSPF Quality of Service Open Shortest Path First) allows routing only in terms of advertised router speed and congestion. In tandem, QOSPF is unable to make the fullest use of CR-LDP as it cannot make use of the detailed traffic descriptions used in CR-LDP; neither can it provide detailed route information. As well as this QOSPF is not able to ensure a connection over a suggested route.

It is accordingly an object of the present invention to provide a method of establishing a connection between two endpoints in a communications network, which overcomes or at least mitigates one or more of the problems noted above.

SUMMARY OF THE INVENTION

Further benefits and advantages of the invention will become apparent from a consideration of the following detailed description given with reference to the accompanying drawings, which specify and show preferred embodiments of the invention.

According to one aspect of the present invention there is provided a method of establishing a connection between two endpoints in a communications network, said communications network comprising a management layer and a physical layer; said physical layer comprising said endpoints and a plurality of nodes interconnected by links and said management layer comprising a plurality of management nodes each management node being connected to a physical node; and wherein said method comprises the steps of: (i) establishing a first path over said management layer between two management nodes, one of said management nodes being connected to one of said endpoints and the other management node being connected to the other end point; and (ii) establishing a second path between said endpoints over said physical layer wherein said first and second paths correspond; and wherein said step of establishing the second path is performed as an integral part of said step of establishing the first path.

A corresponding communications network is provided comprising at least two endpoints between which it is desired to establish a connection, said communications network comprising: (i) a physical layer comprising said endpoints and a plurality of nodes interconnected by links; and (ii) a management layer comprising a plurality of management nodes, each management node being connected to a physical node, and a first one of said management nodes being connected to an endpoint and a second one of said management nodes being connected to the other endpoint; said management layer being arranged to establish a first path between said first and second management nodes; and said physical layer being arranged to establish a second path between said endpoints, over said physical layer, and corresponding to said first path; and wherein said communications network is arranged such that establishment of said second path is an integral part of establishment of said first path.

A corresponding computer program is provided, stored on a computer readable medium, said computer program being for controlling a communications network comprising at least two endpoints, a management layer and a physical layer, said physical layer comprising said endpoints and a plurality of nodes interconnected by links and said management layer comprising a plurality of management nodes each management node being connected to a physical node; said computer program being arranged to control said communications network such that: (i) a first path is established over said management layer between two management nodes, one of said management nodes being connected to one of said endpoints and the other management node being connected to the other endpoint; (ii) a second path is established between said endpoints over said physical layer wherein said first and second paths correspond; and (iii) said establishment of the second path is performed as an integral part of said establishment of the first path.

This provides the advantage that a communications path is established between two endpoints in a communications network quickly such that the set-up latency is low.

In a preferred embodiment a communications session is established which has a guaranteed quality of service. Switch virtual circuit equivalency is effectively given for a communications network which can be an internet protocol based communications network such as an MPLS network.

Preferably, said communications network comprises a plurality of nodes connected together by links and said method further comprises the step of configuring the communications network such that the links between a first plurality of nodes are of a pre-determined capacity such that in use each of said links between the first plurality of nodes is capable of sustaining a plurality of separate communication sessions. By provisioning the communications network in this way high capacity routes which act as "motorways" are created. By using these high capacity routes, the topology information required to implement the method is reduced. This simplifies the method and makes it faster to operate.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram of a communications network.

FIG. 2 is a flow diagram of the process of dynamic label switch path addition.

FIG. 3 is a flow diagram of basic SIP operation with a proxy.

FIG. 4 is a flow diagram showing use of a record-route header to track a route.

FIG. 5 is a flow diagram showing forking with non-explicit abstract nodes.

FIG. 6 is a flow diagram illustrating the process of forming a path element from a record-route header.

FIG. 7 illustrates a basic COPS model.

FIG. 8 is a flow diagram illustrating COPS messaging.

FIG. 9 is a flow diagram illustrating CR-LDP path set-up.

FIG. 10 is a flow diagram illustrating signalling during set-up of a communication session.

FIG. 11 is a schematic diagram illustrating advertisement of labels by label switch routers.

FIG. 12 is a schematic diagram of the process of label propagation.

FIG. 13 shows details of a COPS messaging process.

FIG. 14 shows details of an LDP messaging process.

FIG. 15 is a schematic diagram of the communications network of FIG. 1 with call servers.

FIG. 16 is a schematic diagram of a first method of establishing a bi-directional communication session over the communications network of FIG. 15.

FIG. 17 is a schematic diagram of a second method of establishing a bi-directional communication session over the communications network of FIG. 15.

FIG. 18 is a schematic diagram of a third method of establishing a bi-directional communication session over the communications network of FIG. 15.

FIG. 19 is a schematic diagram of a process for establishing a mapping using the COPS protocol.

FIG. 20 is a schematic diagram of a fourth method of establishing a bi-directional communication session over the communications network of FIG. 15.

DETAILED DESCRIPTION OF THE INVENTION

Embodiments of the present invention are described below by way of example only. These examples represent the best ways of putting the invention into practice that are currently known to the Applicant although they are not the only ways in which this could be achieved.

The term "bidirectional communication session" is used to refer to a period of communication over a communications network that involves transfer of information from a calling party to a called party and vice versa. The bidirectional communication session uses either a single two-way communication path or a forward communication path and a reverse communication path.

Pending U.S. patent application Ser. No. 09/345,069, also assigned to Nortel Networks Corporation, describes a method of establishing a connection between two endpoints in an internet protocol communications network, to provide a guaranteed level of quality of service. The contents of U.S. patent application Ser. No. 09/345,069 are incorporated herein by reference.

Although the method described in U.S. Ser. No. 09/345,069 is in general effective, under some circumstances the time taken to establish such a connection is considerable, for example, for large communications networks such as carrier grade networks. Also, although the method described in U.S. Ser. No. 09/345,069 enables bi-directional communications session to be set-up, the resulting bi-directional communication sessions are not always optimally arranged for some situations.

FIG. 1 is a schematic diagram of a communications network. A first endpoint 10 is connected to another endpoint 11 via a communications network which comprises a plurality of nodes that are connected by links. These nodes include three abstract nodes 12, 13, 14 and many other nodes which are not shown individually but which are represented by cloud shapes 15, 16 between the abstract nodes. These cloud shapes 17, 18 are intended to represent parts of the communications network which in one embodiment is an MPLS network. The cloud shapes 17, 18, nodes 12, 13, 14 and endpoints 10, 11 comprise a physical layer of the communications network.

Links 17, 18 are provided and these connect the abstract nodes 12, 13, 14 in series. Links 19, 20 are also provided to connect each endpoint 10, 11 to an abstract node and thus form a path or tunnel between the endpoints. However, this path from the first endpoint 10, via link 19 to abstract node 12 which is connected in series to abstract nodes 13 and 14, and then via link 20 to the second endpoint 11, is only one of many possible paths over the communications network which connect the two endpoints 10, 11. These other paths are not explicitly shown in FIG. 1 but are intended to be represented by the presence of clouds 15, 16. The links 17, 18, 19, 20 also form part of the physical layer of the communications network.

Data or messages which are transmitted over the communications network can be thought of as comprising two types. First, customer data or messages such as video signals, voice signals or email messages and second, control data or messages. This control data functions to help manage the communications network; for example, control messages may comprise signals broadcast by a node in the communications network to advertise its presence or its failure. The method of using the control messages is defined by the type of messaging protocol(s) used.

In a preferred embodiment of the present invention, the MPLS standard messaging protocol is used in conjunction with the CR-LDP messaging protocol to help manage the communications network comprising the endpoints 10, 11, the abstract nodes 12, 13, 14, the clouds of nodes 15, 16 and the links between these. However, CR-LDP, while able to make quality of service reservations across known paths, is unable to determine these paths itself. In the present invention additional components and messaging protocols are provided, for example, in order to determine and reserve guaranteed quality of service for particular connections for particular paths over the network.

These additional components comprise an administrative server 35, admission managers 30, 31 and connection managers 32, 33, 34 and together these components form a management layer of the communications network. The admission managers 30, 31 and connection managers 32, 33, 34 are referred to collectively as "management nodes". The additional messaging protocols include the standard Common Open Policy Service (COPS) messaging protocol and a modified version of the standard IETF SIP (Session Initiation Protocol) RFC2543 protocol although these are all examples of preferred messaging protocols; any suitable messaging protocols may be used. The modified version of SIP is designed to work in conjunction with COPS, CR-LDP and MPLS, although it could be designed to work with similar messaging protocols to perform the same function. This modified version of SIP will hereinafter be referred to as "SIP++".

In a preferred example, once a connection has been established using the method of the present invention, messages are transmitted over that connection using a protocol that involves labels. Each message or packet contains a header with a label and each abstract node contains a mapping which is a type of routing table. Using a label an abstract node is able to determine which next neighbour abstract node to forward the packet or message to. The mapping details, for received labels, which label should be given to the message or packet when it is forwarded by the abstract node. That is, when a message or packet is received by an abstract node, it identifies the received packet's label in the mapping and determines which label the packet should now be given and hence where the packet should be forwarded next. The packet is then issued with the new label according to the mapping and forwarded by the abstract node. In order for this method to enable messages or packets to be successfully transmitted between two specified endpoints, appropriate label mappings must be set up at the abstract nodes involved. In the method of U.S. Ser. No. 09/345,069 this is achieved after the SIP++ process is complete, using the CR-LDP protocol. That is, the SIP++ process is first used to determine a preferred path between two endpoints which will provide a guaranteed quality of service. Information about this path is then given to the physical layer and using the CR-LDP protocol an actual connection is established along the chosen path. This means that two effectively separate processes occur in series. In the present invention the label mappings are achieved in a different manner which enables the processes of choosing a suitable path and establishing a communication session over that path to be integrated rather than carried out in series.

In one example, Switch Virtual Circuit (SVC) admission control equivalency is provided with guaranteed quality of service on an MPLS or similar communications network. An SVC is a path over a communications network between two endpoints which is effectively dedicated for a particular communication session. These SVCs may be used to carry one or more communication sessions.

In the method of U.S. Ser. No. 09/345,069 two main stages were involved. The first stage involved selection of a path between the two endpoints using the SIP++ protocol. This SIP++ stage took place in the management layer of the communications network and involved propagation of control messages in at least one whole round trip between the two endpoints. Once this stage was completed, the second stage of using the CR-LDP protocol to establish a communication session over the selected path (selected in the SIP++ stage) took place. This second, CR-LDP stage also involved propagation of control messages in at least one whole round trip between the two endpoints. Because of this, the time taken to establish a connection using the method of U.S. Ser. No. 09/345,069 involved at least two round trip delays. As the communications network scales, a single round trip takes a considerable length of time and for carrier grade networks, the delay becomes a real problem when using the method of U.S. Ser. No. 09/345,069.

When a user requests a connection for a communication session this request is passed to an endpoint to which a terminal accessed by the user is connected. Means is provided to determine possible paths for the required connection together with measures of preference for these possible paths. The measures of preference (for example, ranks) are determined on the basis of factors such as traffic levels in the network, length of path, and available capacities. One path is chosen on the basis of the measures of preference. For example, a path with the highest rank may be chosen and reserved for the requested communication session. This gives a reserved path which can be used to provide a guaranteed quality of service for a particular communication session. Any suitable measure of preference such as a score, percentage value or rank may be used.

In an embodiment of the invention a ranking mechanism is used to select from the set of suitable paths, the route a new session will use to traverse an MPLS network. This set of paths and their ranking varies with network load.

In order that the ranks may provide an effective means for choosing between possible paths an advertising mechanism is provided which allows entities in the communications network to gain information about traffic levels, topology of the network and other factors. This information can then be used to help make the decision about which path to choose. The advertisement mechanism allows the system to choose routes best suited to the session being established. Two methods are proposed: explicit registration or by passively piggybacking information on path setup messages. The rate of advertisement is a function of the rate of session set-up.

As well as an advertising mechanism, in order to reduce the complexity of choosing a path, a mechanism is provided whereby an overlay network is configured to provide a set of high capacity routes across the MPLS clouds which function as "trunk" routes or "motorways". An arrangement is then made that communication sessions are preferably established using these pre-determined high capacity routes. This helps to reduce the topology information needed to establish a path across a communications network. By using a constrained set of paths between the routers that comprise the MPLS network, the set of routes is constrained to reduce the total topology information needed to route across the network.

Referring again to FIG. 1, it can be seen that the admission managers 30, 31 and the connection managers 32, 33, 34 as well as the administrative server 35 are depicted above the MPLS network. The admission managers, connection managers and administrative server can be though of as a "management layer" of the communications network. However, this layer is not physically independent from the rest of the communications network. For example, the SIP++ protocol control messages may be transmitted over the same physical links as the user information during communication sessions.

Each endpoint 10, 11 is associated with an admission manager 30, 31 and each abstract node 12, 13, 14 is associated with a connection manager 32, 33, 34. As indicated in FIG. 1, communication between the endpoints and their associated admission managers and between the abstract nodes and their associated connection managers is carried out using the COPS protocol. Also, communication between the administrative server 35 and the admission managers 30, 31 or abstract nodes 12, 13, 14 takes place using the COPS protocol. The way in which this is achieved using the COPS protocol is described in more detail below. However, communication between the admission managers and connection managers takes place using sip++.

The characteristics of some of the components of the communications network are now described:

Abstract Nodes 30, 31

Abstract nodes are a concept introduced by the CR-LDP protocol and represent one or more label switch routers (LSRs) which are connected together by links. By using a description equivalent to a subnet mask a whole group of LSRs can be referred to. A subnet mask is an Internet Protocol (IP) mechanism used to define a group of IP nodes by only using the first n bits of their 32-bit IP addresses, where n is less than 32. The abstract nodes run the CR-LDP protocol and remain unaware of the SIP++ protocol running between admission managers and connection managers. Each abstract node may be directly configured by the Administrative Server, which may instruct an abstract node to establish a path to another particular abstract node. In the case where a CR-LDP network is used this path is referred to as a label switch path (LSP). SIP++ or any other suitable messaging protocol used provides a means of determining which of the label switch routers in an abstract node a path should be routed through.

By using abstract nodes when selecting path candidates for a new session it is possible to be presented with a set of diverse routes. This provides the advantage that different routes over the network can be utilised and this is especially helpful if it is required to "spread load" over the network and if problems occur in localised regions of the network.

Endpoints 10, 11

An endpoint is any node in the communications network through which a user may request a communication session on the communications network. For example, in the case that an MPLS communications network is used an endpoint can be any MPLS device; either an MPLS enabled terminal or a router at the edge of the network. New communication sessions requested by an endpoint are sent to an admission manager that is associated with the endpoint. That admission manager then uses the SIP++ protocol and a path for the requested session is determined and reserved in order to guarantee the requested quality of service. Once the admission manager has completed this task, the user request is validated and the validation communicated to the endpoint using the COPS protocol. Together with the validation, details of the chosen, reserved path are provided to the endpoint together with an identifier for the reserved path. If the request for a new session is granted, the endpoint runs the CR-LDP protocol using the exact same parameters that were used in the COPS request for a communication session together with the details of the chosen, reserved path. The CR-LDP protocol then establishes a path for the communication session according to the standard CR-LDP method described below. Each endpoint is therefore effectively unaware of the SIP++ protocol running between the admission managers and connection managers.

Admission Managers 30, 31

Each admission manager is responsible for maintaining network topology information and using this to select a route across the network. When an admission manager receives a request for a communication session from an endpoint 10, 11 it issues a plurality of path requests, which in a preferred example of the SIP++ protocol are referred to as INVITE messages. These path requests are control messages whose function is to request and determine possible paths between the required endpoints. In order to issue these path requests effectively, an admission manager needs to maintain accurate topological information about at least part of the communications network. Route advertisements are broadcast by entities in the communications network and an admission manager processes all the route advertisements it receives. This enables the admission manager to build up a map of all the reachable nodes on the MPLS network and their availability over time. An admission manager also monitors the bandwidth of connections to edge abstract nodes for the endpoint EP that it is associated with. (An edge abstract node is an abstract node that is positioned towards the edge of a communications network.) In this way an admission manager effectively provides admission control to the communication network. Communication between an admission manager and its associated endpoint is via an interface such as a COPS interface. An interface to the administrative server 35 is also provided, which may be a COPS interface. This allows endpoints to request new tunnels or paths (for example new "trunk" routes) in the communications network such as an MPLS network. An admission manager is also arranged to respond to INVITE messages issued by other admission managers. This is described in more detail below.

Connection Managers

Each connection manager is associated with an abstract node and as described above an abstract node may comprise one or more Label Switch Routers LSRs. However, it is not essential for all label switch routers to be associated with a connection manager.

Connections from these label switch routers to other abstract nodes are termed "label switch paths" (LSPs). Each connection manager monitors the bandwidth used in each of the label switch paths that emanate from the label switch router (or group of label switch routers) which it is associated with (or managing). It also is responsible for advertising the level of congestion in these label switch paths to other administrative elements (such as other connection managers and admission managers) on a slow but regular basis.

A connection manager also keeps a record of the destination abstract node for each of the label switch paths that it is monitoring. This information is also advertised by the connection manager. A connection manager also uses a COPS interface from the abstract node it is monitoring to allow registration of new label switch paths or a change in parameters of an existing label switch path.

Administrative Server

An administrative server 35 is used to provision paths in the communications network upon initialisation. For example, this involves establishing the label switch paths that the SIP++ protocol routes over. It is also used to change the characteristics of an existing path or introduce a new one. Although pictured as a single entity in FIG. 1, an administrative server 35 may take the form of multiple servers that administer their local area.

An Administrative Server is able to communicate directly with any label switch router in a `known` abstract node. It uses CR-LDP over this interface to provision high capacity label switch paths between these label switch routers via any number of intermediate label switch routers. Typically this will be through label switch routers with no associated connection manager, though this need not necessarily be the case. An administrative server has a much more detailed view of the topology of the intermediate MPLS network than the endpoints attached to it. (The intermediate MPLS network being that part of the communications network which is not local to the endpoints.) By pre-provisioning label switch paths of high capacity the administrative server constrains the number of possible routes between two endpoints for a proposed communication session of a given capacity. This reduces the level of detail needed to make routing decisions. Such pre-provisioned label switch paths are referred to as "tunnels".

In a preferred embodiment, when the communications network is first established, it sets up a network of tunnels in the physical layer. These tunnels are subsequently registered with the management layer. That is, information about the source, destination and capacity of each tunnel is made known to the management layer. Each management node makes a record of the tunnels which originate or terminate at the abstract node associated with that management node. These tunnels are each uni-directional. However, in one embodiment, the tunnels are established such that equal sized tunnels exist in either direction between two label switch routers. That is, if an LSR is used in the path to an endpoint, for every tunnel that terminates on that LSR, a corresponding tunnel is provided from that LSR in the opposite direction.

An Administrative Server may also add new paths or change the characteristics of an existing path during the operation of the network. This may either be initiated by the network provider or via a request mechanism which is now described.

Request mechanism

The Administrative Server 35 has a COPS interface to all the admission managers at the edge of the network. This interface is used by those admission managers to request new high capacity label switch paths across the MPLS network, or to request a change in the capacity of an existing LSP.

FIG. 2 shows the process of requesting a new LSP. Either an Endpoint 10 or an Admission Manager 35 issues a Request for a new route between two Abstract Nodes 12, 13 in the MPLS network. This is responded to by the Administrative Server, with the acceptance situation being illustrated in FIG. 2. The Administrative Server now signals to one of the specified abstract nodes AN1, 12 that it should set-up a path to the other abstract node AN2, 13. In the case that the abstract nodes represent a group of label switch routers, the administrative server specifies a particular label switch router within each abstract node.

The first abstract node 12 then registers the requested new path and its characteristics with its Connection Manager 32. This is achieved by issuing a COPS Request message over the COPS interface. The connection manager 32 does not refuse this Request under normal operation and issues a COPS Decision message to this effect. Once a Decision is received by the first abstract node 12, this abstract node proceeds to use CR-LDP to establish the connection to the other specified abstract node. Once the new route is established, the connection manager 32 begins to advertise its presence and the new route can be used immediately in the path for a new session.

SIP++

A simplified SIP++ messaging diagram is provided in FIG. 3, with a brief explanation of the role of each message. These messages are similar to those of SIP but the contents of the messages are modified as compared to SIP. Vertical lines 301 and 302 in FIG. 3 represent two endpoints between which a proxy is located, which is represented a vertical line 303. Messages are sent between these endpoints and the proxy as indicated by the arrows between the vertical lines.

SIP++ registration method

The registration method involves an endpoint, such as endpoint B represented by vertical line 302, sending its internet protocol address to another endpoint, such as end point A represented by vertical line 301.

SIP++ call set-up method

The call-set up method involves an INVITE message being sent from an originating endpoint 301 to the destination endpoint 302. If this INVITE is accepted by the destination endpoint 302 a so called 200 OK message is sent by the destination endpoint 302 to the originating endpoint 301. If the INVITE is not accepted an error response is sent in place of the 200 OK message. Once a 200 OK message is received by an originating endpoint an ACK message is returned to acknowledge receipt of the 200 OK message. This completes the call set-up.

SIP++ Tear Down Method

The tear down method involves either endpoint in a communications path terminating a call by issuing a BYE message to the other endpoint.

SIP++ Request Cancelling Method

This method involves for example, endpoint B 302 starting to make a call to endpoint A 301 and then deciding not to make this call after all. In this situation, endpoint B is able to issue a CANCEL message to endpoint A.

The method of establishing a path for a communication session with a guaranteed quality of service is now described together with an overview of the SIP++ method. Full details of SIP++ are described later.

When a COPS Request is received at an admission manager (requesting a path for a communication session), then providing admission is granted by the admission manager, one or more INVITE messages are sent out by the admission manager. The SIP++ INVITE message extends the standard SIP INVITE message to include a new message body type. Each INVITE message contains a description of the requirements for the desired communication session. For example, the traffic characteristics which are used to establish the path by CR-LDP. A path description is contained within this new body to find a route across the MPLS network that the new session could use. For example, the path description can be a list of nodes which must be visited in sequence to cross the network and reach the required endpoint. Some of the nodes may be unknown and represented as wildcards in the list. Each potential path is also assigned a rank which indicates the admission manager's preference for the route.

For a given INVITE message, the path description is examined and the first reachable abstract node in the list identified. The INVITE message is then sent to the connection manager associated with that reachable abstract node. This is repeated for each INVITE message issued by the admission manager.

When a connection manager receives an INVITE message, it examines the information about the session requirements and next abstract node to see if it has a path to that abstract node and if it can accommodate the new session. There may be more than one path depending on how well defined the abstract node is (for example, if the next abstract node is represented in the path description by a wildcard). If the answer is yes to both questions, it adds the explicit address (such as an IP address) of the abstract node that it is associated with to the INVITE message. An identifier for the connection manager itself is also added to the INVITE message. This information is added to a route-record header field of the INVITE message.

The connection manager then makes a temporary reservation for the session and forwards the INVITE message to the next abstract node in the path description. (If there is more than one abstract node at the next stage of the path description, the INVITE message is "forked" as described below.) If there are insufficient resources or there is no label switch path to the next abstract node in the path description, the connection manager will respond with an error message. This process is repeated until the INVITE messages reach the destination endpoint.

The destination endpoint waits for and collates the incoming INVITE messages. When these INVITE messages were issued by the originating admission manager, they were each assigned a rank by that admission manager. This rank indicates the favourability of a particular path and is scored based on how congested the network appears to the originating admission manager. The rank or other measure of preference is also determined on the basis of factors such as the suitability of the returned path to the type of session being established based on, for example, the latency of the path when establishing a real-time session. The admission manager associated with the receiving endpoint now assigns its own rank to the paths specified in the Record-Route header of each received INVITE message. For each path, the rank from the originating admission manager and from the receiving admission manager is combined in any suitable way, for example by addition, convolution or multiplication. The path and associated INVITE message with the highest scoring rank is then chosen.

As described above, each management node contains a record of any tunnels that originate or terminate at an abstract node associated with that management node. Each management node also advertises these tunnels to each of its next-neighbour management nodes, where a "next-neighbour" management node is one which is directly connected to the management node via a tunnel. Each such advertisement contains information about the source, destination and capacity of the tunnel concerned.

Each of the associated abstract nodes, (i.e. those at which a tunnel originates or terminates) advertises one or more labels that may be used by future communication sessions to traverse the tunnel concerned. These advertisements are made directly to their controlling admission manager to connection manager only, in the management layer. This differs from the method described in U.S. Ser. No. 09/345,069 in which labels were not advertised from the physical layer to the management layer.

When an abstract node advertises a label, information about that label reaches and is stored or cached at the management node that is associated with the "downstream" end of the tunnel. The term, "downstream" is used to refer to a direction along a communication link which is towards the required destination of that communication link. The term, "upstream" is used to refer to the opposite direction. The advertisements are made using the COPS interface between the management and physical layers or any other suitable interface and message protocol. For example, a Label Distribution Protocol (LDP) and interface may be used.

At this stage in the method, the preferred path has not yet been chosen using the SIP++ method and yet labels for use in a communication session are already being advertised. In this way the two processes of path selection and session establishment become integrated.

FIG. 11 illustrates the advertisement of labels by abstract nodes or label switch routers (LSRs). FIG. 11 illustrates a communications network with a physical layer comprising abstract nodes 1200 and endpoints 1201, 1202 in an MPLS communications network. The communications network has been pre-configured as described above and tunnels are provided in the direction indicated by arrow A. A communication session is required to be established between endpoint 1202 and endpoint 1201 in the direction of arrow A. This direction is termed "downstream".

As described above the first stage of the SIP++ negotiation process takes place as in U.S. Ser. No. 09/345,069. That is, INVITE messages are sent from an admission manager 1203 associated with the first endpoint 1202, to the admission manager 1204 associated with the second endpoint 1201. The next stage of the SIP++ process involves sending back a 200 OK response from the second or destination endpoint 1201 to the originating endpoint 1202. During this 200 OK response stage, advertised labels are chosen in such a way that a label mapping is established at each LSR. Then once the 200 OK response reaches the originating admission manager 1203, label mappings have been established at each LSR along the route. Because the labels are selected during the 200 OK response stage, which is an upstream process, the labels have to be selected in such a manner that they can later be used for the required downstream communication session. This is achieved as described below.

When an abstract node advertises a label (see 1205 in FIG. 11), information about that label reaches and is stored or cached at the management node 1204 that is associated with the "downstream" end of the tunnel. For example, as shown in FIG. 11, labels advertised by endpoint 1201 are stored at admission manager 1204 rather than at connection manager 1206.

The receiving admission manager now forms a 200 OK response to the chosen INVITE message. The 200 OK response needs to be returned along the same path as the chosen INVITE message arrived. The path along which the chosen INVITE message arrived is known from the details of each abstract node passed on route. This information is taken from the Record-Route header of the chosen INVITE message and used to form a new path description for the 200 OK message. Also, the Record-Route header of the chosen INVITE message is copied into the 200 OK message. The 200 OK message is then sent back to the originating admission manager.

However, before the 200 OK response is sent back, the admission manager 1204 (FIG. 11) consumes a label for the communication session. That is the admission 1204 manager selects one of the advertised labels that are stored or cached at that admission manager. The admission manager then informs its associated endpoint 1201 that it has chosen a particular label, in order that the endpoint 1201 does not advertise that label as being available any more. The admission manager also informs its associated endpoint 1201 about the identity of the required communication session. The endpoint then knows that the chosen label and the required communication session are associated.

The admission manager 1204 now sends the 200 OK response together with the chosen label, to the first connection manager 1206 in the record-route list. That connection manager 1206 then converts its temporary reservation for the requested communication session into a permanent reservation. Next, the connection manager 1206 examines the record route list to identify which connection manager to forward the 200 OK response to. Supposing that this next connection manager is 1207 in FIG. 2, the first connection manager 1206 determines a tunnel to connection manager 1207 and consumes a label for that tunnel. That is, one of the labels cached at the first connection manager 1206, which is suitable for the chosen tunnel, is selected.

Information about this chosen label is sent from the first connection manager 1206 to its associated abstract node in order that that abstract node does not re-advertise the particular label. Also, information about the previously chosen label (that suitable for the journey between the first connection manager 1206 and the destination endpoint 1201) is sent from the first connection manager 1206 to its associated abstract node. This enables the associated abstract node to set up a label mapping. In future, if the associated abstract node receives a packet with a label corresponding to the most recently chosen label, it "knows" to forward the packet to the destination endpoint using the previously chosen label.

The first connection manager 1206 then sends the 200 OK response to the next connection manager together with the most recently chosen label and the process repeats until the 200 OK response reaches the originating admission manager 1203. This process is illustrated in FIG. 12 which shows a 200 OK response 1300 carrying a label between the destination admission manager and the first connection manager 1206, and also between subsequent connection managers 1301, 1302 in a path to the originating admission manager 1203. Communication between the physical and management layers is also illustrated. Decisions about which label to select are sent from the management layer to the physical layer and information about corresponding pairs of labels is communicated 1304 between next-neighbour abstract nodes in the physical layer.

When the 200 OK response reaches the originating admission manager 1203, the originating admission manager 1203 informs the originating endpoint 1202 which label it should use to reach the first abstract node (or LSR) in the selected path. This information is matched up with a session description that the originating endpoint holds for the forward mapping of the required communication session. At this point the path between the endpoints 1202, 1201 is completely specified and established for the required communication session. This is achieved without the need for a separate CR-LDP negotiation to take place, after the SIP++ negotiation is complete. Hence, one complete round trip of control messages is eliminated and this significantly reduces session setup time.

FIG. 13 illustrates the details of the COPS messaging process for the situation where the registration of tunnels is already complete and advertised labels have been cached. In FIG. 13 the management nodes and physical nodes are represented by vertical lines with pairs of management nodes and associated physical nodes next to one another. The downstream direction is indicated by arrow B.

One cached label 1400, 1401, 1402 is shown for each management node 1404, 1404, 1405 but this is for illustrative purposes only. A plurality of labels may be cached at each management node.

An INVITE message 1406 is illustrated as being rec