PointtoMultipoint in 802 1 Qay Nurit Sprecher Nokia

Point-to-Multipoint in 802. 1 Qay Nurit Sprecher, Nokia Siemens Networks Hayim Porat, Ethos Networks Slide 1 11 May 2008

Background • The spanning tree is disabled in PBB-TE regions. – Restoration is disabled when a failure occurs along a TE service instance. – Protection switching is enabled instead, but only for point-to-point TE service instances. – In the event of a failure along a point-to-multipoint TE service instance, traffic is not restored until the failure is eliminated (hours? days? ). • In PBB-TE, no control protocol is enabled; configuration is controlled by an external agent. Only protection switching can be applied to point-to-multipoint TE service instances (as restoration requires control protocols). Slide 2 11 May 2008

Observations • In contrast to point-to-point TE service instances, it is difficult to set up point-to-multipoint TE service instances without knowledge of: – the point-to-multipoint customer services which are to be delivered over the PBB-TE region with their corresponding source CBP and destination CBPs. • This is due to the extremely high number of possible point-to-multipoint TE service instances that can be setup between N CBPs in a PBB-TE region. – The number of possible point-to-multipoint TE service instances consisting of 2 and above destination CBPs in a PBB-TE region consisting of N CPBs is ( 2 N– 1 Slide 3 – N ) * N. 11 May 2008

Different options for point-to-multipoint TE service instances in 802. 1 Qay 1. One point-to-multipoint TE service instance in the PBB-TE region per point-to-multipoint customer service (which does not scale) 2. Using a point-to-multipoint TE service instance for a group of pointto-multipoint customer services which may result either in inefficient link utilization (traffic is replicated to all branches), or in one point-tomultipoint TE service per point-to-multipoint customer service in the PBB-TE region. Slide 4 11 May 2008

Point-to-multipoint TE service instance per group of point-to-multipoint customer services Inefficient BW Utilization A Two Multicast Groups: ABD, AC B Distributed to all branches Or C D Slide 5 11 May 2008

Point-to-multipoint TE service instance per group of point-to-multipoint customer services Point-to-multipoint TE Service Instance per customer service A Two Multicast Groups: ABD, AC B Distributed to all branches Or Split based on MAC address C D Slide 6 11 May 2008

Different options for protection switching • Segment-based protection (e. g. per branch): – The protection entity may be point-to-point or point-to-multipoint TE service instance. – Requires the definition of MEPs on Network Provider ports of core bridges in a PBB-TE region – Requires ESP-VID translation inside the PBB-TE region when protection switching occurs. Violates the scope of the domain-wide label of PBB-TE – May be a complex problem when there are sub-branches of sub-branches • 1: 1 point-to-multipoint protection: To guarantee protection, disjoint point-to-multipoint TE service instances should be set-up. ▪ This consumes a lot of resources, especially since, in practice, one point-to-multipoint TE service instance is allocated per point-to-multipoint customer service. ▪ This puts a lot of constraints on the resource allocation algorithm, which may result in the inability to provision protected point-to-multipoint TE services. Slide 7 11 May 2008

Different options for protection switching (Contd. ) • 1: 1 point-to-point protection: 1. To guarantee protection, disjoint point-to-point TE service instance should be set-up for each destination CBP. ▪ When a failure occurs along a particular branch from the source CBP to a destination CBP, a switchover is performed from the failed branch to the relevant protection point-to-point TE service instance, without affecting the entire point-to-multipoint TE service insatnce. ▪ Although in practice one point-to-multipoint TE service instance is allocated per point-tomultipoint customer service, this result in a better resource utilization as the protection point-to -point TE service instances for the different destination CBPs do not need to disjoint. ▪ This does not put additional constraints on the resource allocation algorithm as compare to the 1: 1 point-to-point protection switching. 2. Point-to-multipoint customer services are delivered over full-mesh protected point-topoint TE service instances. ▪ This does not require any modification to the protection switching mechanism which is defined in 802. 1 Qay. Slide 8 11 May 2008

Recommendations Since point-to-multipoint TE service instances should be adequately protected, 1: 1 protection switching should be defined. It is recommended to adopt the 1: 1 point-to-point protection switching either by the – Pre-configuration of a disjoint point-to-point TE service instance for each destination CBP, or by the – Delivery of point-to-multipoint customer services over full mesh protected point-to-point TE service instances. Slide 9 11 May 2008

Thank You Slide 10 11 May 2008

Backup Slides Slide 11 11 May 2008

Protection vs. Restoration • Protection (data plane) – Utilizes the pre-assigned capacity between nodes. – 50 ms is the target time for protection switching. – In the event of failure, the protection mechanism does not involve rerouting or an additional connection setup. Only the source and destination connection controllers are involved. – Protection switching may be either unidirectional or bidirectional. – Protection at link, trunk or circuit layers – Optimization for specific topologies: Linear Protection, Ring Protection • Restoration (distributed control plane, management plane) – Utilizes any available capacity between nodes – The algorithms used for restoration involve rerouting. When restoration is used, a percentage of the transport network capacity is reserved for the rerouting of working traffic. – Any topology. Slide 12 11 May 2008

Segment-based Protection (e. g. per branch) MEPs need to be defined to monitor the healthy of the segments ESP-VID translation is needed A B Note: for simplicity only one protection entity is presented, but protection entities need to be set-up for each segment to guarantee protection switching for the entire point-tomultipoint TE service instance. Slide 13 11 May 2008 C D

1: 1 point-to-point protection with disjoint point-topoint TE service instance for each destination CBP A B Note: for simplicity only some protection point-to-point TE service instances are presented, but protection point-to-point TE service instances need to be set-up from the source BCP to each destination BCP. Slide 14 11 May 2008 C D

1: 1 point-to-point protection with Point-tomultipoint service delivery over full-mesh point-topoint TE service instances Full-mesh protected point-to-point TE service instances. For simplicity only one protection point-to-point TE service instances are presented, but protection point-topoint TE service instances need to be set-up for each point-to-point TE service instance. Slide 15 11 May 2008