Resiliency in minimum cost tree-based VPLS architecture
A system for providing resilient multimedia broadcasting services over a VPLS network is described. A Network Management System (NMS) calculates disjoint minimum cost trees using the Steiner algorithm, executed with extra steps to result in disjoint trees. Destination PE routers in the VPLS network are connected to the disjoint trees so that they can be serviced by either tree in case of a fault. Each of the disjoint trees is provisioned with enough bandwidth to carry all of the services provided by the VPLS network. Under normal operation, however, the services are distributed evenly over the trees. In the event of a fault, the services on a faulty tree are switched to another tree using split horizon bridging. Each Steiner tree can also be realized using point-to-multipoint LSPs which is fully protected by a precomputed point-to-multipoint LSP.
1. A method of providing broadcast services over a Virtual Private LAN Services (VPLS) network having at least two source Provider Edge (PE) routers, each of which is connected to a respective head-end system for broadcasting services, and having a plurality of destination PE routers, each of which is connected to at least two core provider (P) routers in the VPLS network, the method comprising:
calculating a plurality of Steiner trees, each Steiner tree in the plurality of Steiner trees being sourced by a respective source PE router and including each destination PE router from the plurality of destination PE routers, the plurality of Steiner trees being disjoint from one another with respect to at least the core P routers in the VPLS network, said calculating comprising:
computing a first Steiner tree coupled to a first multicast router and having a set of the core P routers,
computing a second Steiner tree coupled to a second multicast router without involving the set of the core P routers of the first Steiner tree in the computations, and
provisioning both the first Steiner tree and the second Steiner tree with enough bandwidth to carry all of the broadcast services but using load balancing so that the first Steiner tree has about half of the bandwidth and the second Steiner tree has about half of the bandwidth;
broadcasting the broadcast services evenly over the first Steiner tree and the second Steiner tree, whereby all of the broadcast services are broadcasted over the first Steiner tree and the second Steiner tree; and
after the first Steiner tree fails, switching all of the bandwidth to the second Steiner tree, wherein the switching step comprises:
inserting static Internet Group Management Protocol (IGMP) memberships in the second multicast router, and
removing IGMP entries from the first multicast router.
2. The method of claim 1 , further comprising:
calculating the first Steiner tree and the second Steiner tree using point-to-multipoint Label Switched Paths (LSPs).
3. The method of claim 1 , further comprising:
detecting a fault in the first Steiner tree to designate a faulty tree and a non-faulty tree, and, in response to detecting the fault; and
attaching, according to a split-horizon bridging, Customer Edge (CE) routers of the faulty tree to the non-faulty tree.
4. The method of claim 3 , further comprising:
freeing bandwidth on the faulty tree.
5. The method of claim 1 , wherein the broadcast services are digital multimedia services.
6. The method of claim 5 , wherein the digital multimedia services are digital television services.
7. The method of claim 5 , wherein the digital multimedia services are video on demand services.
8. The method of claim 1 , further comprising:
using a Network Management system (NMS) to provide the broadcast services.
9. The method of claim 8 , further comprising:
using the NMS to pre-compute the first Steiner tree and the second Steiner tree such that bandwidth is reserved for the first Steiner tree and the second Steiner tree.
10. A Network Management System (NMS) that provides broadcast services over a Virtual Private LAN Services (VPLS) network having at least two source Provider Edge (PE) routers, each of which is connected to a respective head-end system for broadcasting services, and having a plurality of destination PE routers, each of which is connected to at least two core provider (P) routers in the VPLS network, the NMS comprising:
means for calculating a plurality of Steiner trees, each Steiner tree in the plurality of Steiner trees being sourced by a respective source PE router and including each destination PE router from the plurality of destination PE routers, the plurality of Steiner trees being disjoint from one another with respect to at least the core P routers in the VPLS network, wherein said calculating further comprises:
computing a first Steiner tree coupled to a first multicast router and having a set of the core P routers,
computing a second Steiner tree led to a second multicast router without involving the set of the core P routers of the first Steiner tree in the computations, and
provisioning both the first Steiner tree and the second Steiner tree with enough bandwidth to carry all of the broadcast services but using load balancing so that the first Steiner tree has about half of the bandwidth and the second Steiner tree has about half of the bandwidth;
means for broadcasting the broadcast services evenly over the first Steiner tree and the second Steiner tree, whereby all of the broadcast services are broadcasted over the first Steiner tree and the second Steiner tree; and
means for, after the first Steiner tree fails, switching all of the bandwidth to the second Steiner tree, wherein the means for switching comprises:
means for inserting static Internet Group Management Protocol (IGMP) memberships in the second multicast router, and
means for removing IGMP entries from the first multicast router.
11. The NMS of claim 10 , wherein the provisioning for both the first Steiner tree and the second Steiner tree uses point-to-multipoint Label Switched Paths (LSPs).
12. The NMS of claim 10 , wherein split horizon bridging is used to distribute services.