PROTECTION SCHEME

This invention relates to a protection scheme for multicast communication in a network,

and has particular, but not exclusive, application in a network using multi-protocol label

switching (MPLS) for routing packets through the network.

Networks based in a metropolitan area are often referred to as Metro networks. Such

networks are arranged to provide communication capabilities between residential users, business users, Internet service providers, network operators and the like.

Metro networks can advantageously employ MPLS to direct the communication in the

form of packets through the network to an end-user. In these networks there is a demand for communications to be multicast to the end users, for example, in multimedia applications such as video streaming, Internet protocol television (IPTV) or the like.

This is achieved by using multicast label switched paths (LSPs).

For data being streamed to the user for real-time processing, it is necessary for data to be

streamed to the user at a required data rate. Faults in the network can be a barrier to this

requirement, causing traffic interruptions or degradations. To maintain communication at

the required data rate between the supplier and the end-user, even when a fault occurs in

the network, protection schemes are provided in such networks.

Several schemes presently exist which offer restoration and/or protection in the event of a

failure of a multicast LSP within the network.

Figure 1 shows a metro network having one-to-one LSP protection. The network

comprises nodes 2A to 2F connected together by a communication link 1 in the form of a

fibre optic ring. Each egress node 2B to 2F is connected to the final users (in a

residential environment in the figure) via an access node 6. Node 2 A is an ingress node

connected to a video server 3 and an Internet service provider (ISP) 4 through a router 5.

The node 2A is arranged to determine which LSP to send packets on dependent upon the

egress nodes 2B to 2F the packets need to reach, as is conventional in a LSP network.

In Figure 1 there is shown a primary LSP 7 on which packets of a video stream are sent multicast to nodes 2B, 2C and 2E. For primary LSP 7 there is also provided one back-up

LSP 8. The ingress node 2 A is pre-configured with these LSPs 7,8. During normal operation, the packets are sent multicast along the primary LSP 7, however when a fault

is reported to the ingress node 2 A that prevents communication along the primary LSP 7, the ingress node switches the communication to the back-up LSP 8.

This protection scheme has the advantage that it consumes no extra bandwidth for protection compared to other schemes, however, as now illustrated with respect to Figure

2, as the LSPs 7 and 8 are pre-configured prior to failure, in some situations the scheme

can fail to maintain the required multicast communication on the network, thus failing in

practice to reach one or more egress nodes.

Figure 2 shows a fault 10 occurring in the network that prevents communication between

egress nodes 2B and 2C. This fault interrupts communication along the primary LSP 7

to nodes 2C and 2E. The ingress node 2A therefore, on receiving a fault notification 9,

switches communication to the back-up LSP 8. However, this is not sufficient to

maintain the required multicast communication because communications along the back¬

up LSP 8 fail to reach node 2B due to the fault. Therefore, the one-to-one protection

scheme with pre-configured LSPs can fail to maintain the multicast communication.

A scheme that has been developed to overcome this deficiency is shown in Figure 3. In

this scheme rather than the ingress node 2A being pre-configured with a back-up LSP 8,

the ingress node 2 A computes the back-up LSP 8' on-demand in response to a fault

notification 9 that includes information on the location of the fault. The ingress node 2 A

determines a back-up LSP 8 ^* that avoids communication along the optic fibre that has failed. However, in order to achieve a multicast LSP 8' on-demand, the LSP 8' first has

to be computed by the ingress node 2A and then established through signalling between the nodes 2. This has a number of disadvantages. Firstly, the extra computing and signalling that needs to be carried out consumes processing power and to some extent

bandwidth reducing the speed of the network. Secondly, it cannot be easily incorporated into current metro networks because no complete solution for signalling multicast LSPs actually has been standardised to date.

According to a first aspect of the invention there is provided a method of sending a

multicast communication across a network comprising an ingress node connected to a

plurality of egress nodes via a communication link such that communications can be sent

to each egress node from the ingress node along more than one path, the method

comprising configuring the nodes in advance of a fault occurring in the network with a

primary path along the link on which to send multicast communications and a back-up

path along the link on which to send the multicast communications and, if the fault

occurs, sending a multicast communication on the primary path and a duplication of the

multicast communication on the back-up path.

Sending a multicast communication on the primary path and a duplication of the multicast

communication on the back-up path when a fault occurs ensures that the multicast

communication reaches all of the required egress nodes without requiring additional

computing by the ingress node that could significantly reduce the speed of the network protection or signalling to establish a back-up path.

The nodes may be configured in advance of a fault with a plurality of primary paths, each primary path for sending a multicast communication along the link to a unique set of

egress nodes. The nodes may also be configured with a back-up path for each primary

path. In this way, one-to-one protection is provided.

The primary path and back-up path may be multicast label switched paths (multicast

LSPs).

The method may comprise stopping the sending of the duplicate multicast communication

on the back-up path in response to the ingress node being notified that there is no longer

the fault in the network. The response may not be immediate, but may be delayed by a

pre-set amount. In this way, the method ensures that all egress nodes have reverted to receiving the multicast communication via the primary path before switching off the

back-up path. The preset delay may be configured by the network operator, and, for

instance, can range from some milliseconds to a few seconds or even more, depending on

the specific choices of the network operator. Such delay has to be set in order to provide

sufficient time for all egress nodes to revert to using the primary path while avoiding any

significant effects on the data traffic.

The method may comprise switching the egress nodes to revert back to receiving the

multicast communication on the primary path a predetermined period of time after

receiving notification that the fault has been fixed. In this way, the egress nodes only switch back to receiving the communication on the primary path once a stable

communication has been established.

According to a second aspect of the invention there is provided an ingress node for a

network in which the ingress node is connected to a plurality of egress nodes via a

communications link such that communications can be sent to each egress node from the

ingress node along more than one path, the ingress node arranged to be configured in advance of a fault occurring in the network with a primary path along the link on which to send multicast communications and a back-up path along the link on which to send the

multicast communications and to send a multicast communication on the primary path and a duplication of the multicast communication on the back-up path if the fault occurs.

According to a third aspect of the invention there is provided, a data carrier for a

network comprising an ingress node connected to a plurality of egress nodes via a

communications link such that communications can be sent to each egress node from the

ingress node along more than one path and the nodes being configured in advance of a fault occurring in the network with a primary path along the link on which to send

multicast communications and a back-up path along the link on which to send the

multicast communications, the data carrier comprising instructions that when executed by

a processor cause the processor to operate the ingress node of the network such that, in

response to receiving a notification that the fault has occurred, the ingress node sends a

multicast communication on the primary path and a duplication of the multicast

communication on the back-up path.

According to a fourth aspect of the invention there is provided an egress node for a

network in which the egress node is one of many egress nodes connected to an ingress

node via a communications link such that communications can be sent to each egress

node from the ingress node along more than one path, the egress node arranged to be

configured, in advance of a fault occurring, with a primary path on which the egress node can receive multicast communications and a back-up path on which the egress node can receive the multicast communications, to receive a multicast communication along a

primary path and, in response to detecting that communication to the egress node on the primary path has failed, identifying if a duplicate of the multicast communication can be received on the back-up path and, if so, switching to receive the duplicate of the multicast

communication.

It will be understood that "receive" used herein means to detect or to pick up not simply

to have sent to.

The egress node of the invention is advantageous as it only switches to the back-up path if communication in the primary path has failed, independent of each other egress node in

the network. In this way, the ingress node can send communications to the egress node

along the primary path if communications along the primary path are still possible.

The egress node may be arranged to switch to receiving the multicast communication on

the primary path a predetermined period of time after receiving notification that the fault

has been fixed.

According to a fifth aspect of the invention there is provided, a data carrier for a network

comprising an ingress node connected to a plurality of egress nodes via a communications link such that a communication can be sent to each egress node from the ingress node

along more than one path and the nodes being configured in advance of a fault occurring in the network with a primary path along the link on which to send multicast

communications and a back-up path along the link on which to send the multicast

communications, the data carrier comprising instructions that when executed by a processor cause the processor to operate the egress node of the network to receive a

multicast communication along a primary path and, if the egress node fails to receive the multicast communication along the primary path, identifying if a duplicate of the multicast communication can be received on the back-up path and, if so, switching to

receive the duplicate of the multicast communication.

According to a sixth aspect of the invention there is provided a network comprising an

ingress node connected to a plurality of egress nodes via a communications link such that

communications can be sent to each egress node from the ingress node along more than

one path, the nodes configured in advance of a fault occurring in the network with a

primary path along the link on which to send multicast communications and a back-up path along the link on which to send the multicast communications and the nodes

arranged to send a multicast communication on the primary path and a duplication of the multicast communication on the back-up path, if the fault occurs.

The communication link may be a ring, in particular a fibre optic ring, connecting the

nodes together, wherein communication can occur in both directions along the ring. In

this way, two paths are provided to each egress node from the ingress node.

The communication link may provide one-to-one protection for each path through the

network; that is that for each primary path through the network there is a separate back¬

up path.

Alternatively, the communication link may provide many-to-one protection, with multiple

primary paths being protected via a single back-up path.

The communication link may provide protection against failure of the link between two nodes, a node and/or a failure of the link between two nodes as well as a node.

Each egress node may be arranged to receive the duplication of the multicast communication along the back-up path only if the egress node fails to receive the

multicast communication along the primary path. For example, the egress nodes may

operate by switching to receive multicast communications on the back-up path in

response to failure to receive the multicast communication on the primary path.

The ingress node may be arranged to send the multicast communication on the primary

path and a duplication of the multicast communication on the back-up path on receiving a

communication notifying the ingress node of the fault in the network.

The node may be a switch, a router or other network device capable of being deployed in

a packet switched network, conveniently based upon a connection oriented technology.

The network may use multi-protocol label switching (MPLS) for forwarding packets

through the network, with the primary path and back-up path being specific label

switched paths (LSPs) in the network.

The network may use connection oriented Ethernet for forwarding packets through the network, with the primary path and back-up path being specific connections in the network.

The multicast communication may be a communication of data for multimedia

applications or real-time processing at a destination device, for example, the multicast communication may be video streaming or Internet protocol television (IPTV) or the like.

An embodiment of the invention will now be described, by example only, with reference

to the accompanying drawings, in which :-

Figure 1 is a schematic view of a metropolitan network that uses multi-protocol

label switching (MPLS) with pre-established LSP one-to-one protection;

Figure 2 is a schematic view of the network of Figure 1 in which a fault has

occurred;

Figure 3 is a schematic view of a metropolitan network that uses multi-protocol

label switching (MPLS) with LSP one-to-one protection provided on-demand; and

Figure 4 is a schematic view of a metropolitan network that uses multi-protocol

label switching (MPLS) with pre-established LSP one-to-one protection that operates in accordance with the invention.

Referring to Figure 1, a network in accordance with the invention comprises nodes 102 A to 102F connected together by a fibre optic communication link 101 having a ring

topology. Each egress node 102B to 102F is connected to residential networks via an access node 106. Node 102 A is an ingress node connected to a video server 103 and an Internet service provider (ISP) 104 through a router 105.

The network is a MPLS network in which ingress node 102 A is arranged to determine which LSP to send data packets on dependent upon the egress node or nodes 102B to

102F the packets need to reach. The ingress node 102 A then attaches a label to each data

packet, the label being used by downstream nodes 102B to 102F to determine what to do

with the packet.

The ingress node 102A and egress nodes 102B to 102F are configured with both the

primary LSPs on which communications are sent in normal operation of the network and

back-up LSPs on which communications are sent if a fault occurs in the network. In this embodiment of the invention, each primary LSP is provided with a back-up LSP to

provide one-to-one protection.

Figure 4 illustrates with arrow 107 a primary LSP on which packets of a multimedia stream are sent multicast to nodes 102B, 102C and 102E and arrow 108 shows a back-up LSP for that primary LSP.

In the invention, a multicast communication is the sending of data packets from the

ingress node to more than one egress node in the network at substantially the same time,

wherein each fibre optic connection of the network only carries one copy of the

communication, copies of the communication only being made when connections to destination nodes split, for example, in Figure 4 at the egress nodes. This is in contrast to

broadcast or multicast supported by mere packet replication in every point of the network, wherein separate communications are sent to each destination such that any one

link of the network may carry more than one copy of the communication, or unicast, wherein a communication is sent to a single destination. Multicast communications are highly desirable for the communications of data for multimedia applications or real-time processing, such as live media events and the like, to the end-user.

The ingress node 102 A is programmed to send communications solely on the primary LSP unless the ingress node 102A receives a notification from an egress node 102B to 102F of a fault. Fault detection and notification may be based on prior art techniques, such as physical criteria (for detection), or proprietary or standard Operation, Administration and Maintenance (OAM) messages. In response to receiving a fault notification, for communications to the node or nodes that have reported the fault, the ingress node 102 A keeps on sending the communications on the primary LSP 107 but also sends a duplicate of these communications on the back-up LSP 108.

The ingress node 102 A is further programmed to stop sending the duplicate communication along the back-up LSP 108 in response to receiving a notification that the fault has been fixed. The ingress node 102 A may stop sending the duplicate communication a preset time after receiving the notification. This pre-set time may be configured by the network operator according to specific requirements of the network.

The egress nodes 102B to 102F are programmed to initially (and, preferentially) attempt to receive communications sent on primary LSP 107. However, if the egress node 102B to 102F fails to receive a communication on the primary LSP 107, the egress node 102B to 102F will look to see if a valid duplicate of the communication is being sent on the back-up LSP 108. If the egress node 102B to 102F has been receiving a communication on the back-up LSP 108 then the egress node 102B to 102F will revert to using the primary LSP 107 once a stable communication along the primary LSP 107 becomes available. In order to make sure that stable communication along the primary LSP 107 is available again, each egress node 102B to 102F that switched to the backup LSP during a fault will wait for a pre-determined amount of time after receiving notification that the fault has been fixed before reverting back to using the primary LSP 107. This predetermined time is less (preferably much less) than the pre-set time the ingress node 102 A waits before stopping sending the duplicate multicast communication on the backup LSP 108.

As an illustration, a particular example of the networks operation will now be described. Figure 4 shows the network in which a fault has occurred on a portion of the

communication link preventing communication between nodes 102B and 102C. This

fault prevents multicast communication along the primary LSP 107 to nodes 102C and

102E. Such a failure may be caused for example by a mechanical fault in the optical fibre (i.e. fibre/cable cut).

In response to the fault, egress nodes 102C and/or 102E (or also 102B, depending on the

fault detection and notification scheme which is in place) send a notification to ingress

node 102A notifying the ingress node 102A that a fault occurred on the primary LSP 107. In response to this notification, the ingress node 102A sends the communication on

the primary LSP 107 and a duplicate of the communication on the back-up LSP 108.

The egress node 102B can still receive the multicast communication on the primary LSP 107 and therefore, continues to do so. However, the egress nodes 102C and 102E can no longer receive the multicast communication on the primary LSP 107 and respond by

identifying whether they can receive a duplicate of the multicast communication on the

back-up LSP 108. The fault does not prevent communication to nodes 102C and 102E along the back-up LSP 108 and therefore, nodes 102C and 102E switch to receive the valid duplicate multicast communication sent on the back-up LSP 108.

As a duplication of a multicast communication is sent on occurrence of a fault, the

bandwidth of the communication link 101 available for other communications sharing a

common path with the back-up LSP 108 is reduced. Therefore, the data rate for

communications having a lower priority than the multicast communication may be

reduced for the duration of the fault. These considerations are part of normal network planning and dimensioning.

Once the fault is fixed, the egress nodes 102C and 102E identify that communications can now be received on the primary LSP 107 and respond by reverting back to receiving

communications on the primary LSP 107. The ingress node 102A, responds to receiving

notification that the fault has been fixed by stopping sending the duplicate multicast

communication on the back-up LSP 108. The ingress node 102 A may be programmed to

delay stopping the sending of the duplicate multicast communication on the back-up LSP

108 by configurable time. This gives any of the egress nodes 102B to 102F that have

switched to receive the duplicate multicast communication on the back-up LSP 108 sufficient time to revert back to receiving the multicast communication on the primary

LSP 107 (if required by waiting a minimal amount of time, which may be configured, in order to make sure that the primary LSP is stable way) .

This embodiment of the invention is advantageous as the fault does not prevent the communication being received by the appropriate egress nodes 102B to 102F, while avoiding the need for significant additional computing to be carried out by the ingress node 102 A and signalling to establish a LSP on-demand when the fault occurs. Also, contrary to 1 + 1 protection schemes, no extra bandwidth is consumed for protection

during normal operation. In this way, existing networks can be easily adapted to be in

accordance with the invention.

It will be understood that the invention is not limited to the above-described embodiment

but modifications and alterations to the embodiment are possible without departing from

the scope of the invention as defined in the claims. For example, the network may have a tree, mesh, or fully connected topology.

It will be readily appreciated that technologies may also be used other than MPLS, for example connection oriented Ethernet.

It will be understood that the network may be arranged such that any one of the nodes

may act as an ingress node or an egress node dependent on the source of the

data/application to be communicated. Accordingly, in one embodiment, the egress node

for one or more applications is the ingress node for other applications.