Abstract: In one example, a method includes by a first network device positioned on a border of a first area of a multi-area hierarchical network and a second area of the multi-area hierarchical network, determining a cost associated with sending network traffic from a client group to the first network device, wherein the client group is positioned in the first area, the first area and the second area being distinct routing domains of the multi-area hierarchical network; and outputting, by the first network device to a second network device positioned in the second area, a routing advertisement that specifies the determined cost as a reverse metric. In some examples, a route reflector receives the routing advertisement and based on the cost from the client group to the area border network device, selects an egress point from among a plurality of egress points of the multi-area hierarchical network.

Abstract: In general, techniques described are for bandwidth sharing between resource reservation protocol label switched paths (LSPs) and non-resource reservation protocol LSPs. For example, in networks where resource reservation protocol LSPs and non-resource reservation protocol LSPs co-exist within the same domain, resource reservation protocol LSPs and non-resource reservation protocol LSPs may share link bandwidth. However, when non-resource reservation protocol LSPs are provisioned, resource reservation protocol path computation elements computing resource reservation protocol paths may not account for non-resource reservation protocol LSP bandwidth utilization. The techniques described herein provide a mechanism for automatically updating traffic engineering database (TED) information about resource reservation protocol LSPs in a way that accounts for non-resource reservation protocol LSP traffic flow statistics, such as bandwidth utilization.

Abstract: In general, techniques are described for conflict resolution in source packet routing in networking. For example, a first router receives a first advertisement originated in a first Interior Gateway Protocol (IGP) level. The first advertisement specifies a first prefix and a segment identifier (SID). The first router also receives a second advertisement originated in a second IGP level of the network. The second advertisement specifies a second prefix and the SID. Based on the first advertisement and the second advertisement specifying the same SID and based on the first IGP level having less visibility than the second IGP level, the first router selects the SID to be associated with a route to the first prefix.

Abstract: In general, techniques described are for bandwidth sharing between resource reservation protocol label switched paths (LSPs) and non-resource reservation protocol LSPs. For example, in networks where resource reservation protocol LSPs and non-resource reservation protocol LSPs co-exist within the same domain, resource reservation protocol LSPs and non-resource reservation protocol LSPs may share link bandwidth. However, when non-resource reservation protocol LSPs are provisioned, resource reservation protocol path computation elements computing resource reservation protocol paths may not account for non-resource reservation protocol LSP bandwidth utilization. The techniques described herein provide a mechanism for automatically updating traffic engineering database (TED) information about resource reservation protocol LSPs in a way that accounts for non-resource reservation protocol LSP traffic flow statistics, such as bandwidth utilization.

Abstract: An example method includes selecting, by a network device, a remote LFA next hop as an alternate next hop for forwarding network traffic from the network device to a destination, wherein the selected remote LFA next hop provides node protection to a primary next hop node on the shortest path from the network device to the destination. The method includes, for each candidate remote LFA next hop, performing a forward shortest path first (SPF) computation having the respective candidate remote LFA next hop as a root to compute a path segment between the respective candidate remote LFA next hop and the destination, wherein each of the candidate remote LFA next hops is the egress of a respective potential repair tunnel between the network device and candidate remote LFA next hop, and selecting the remote LFA next hop based at least in part on the computed path segments.

Abstract: In general, techniques are described by which to provide a scaled end-to-end view of link metrics to integrate multiple non-uniform Interior Gateway Protocol (“IGP”) domains. For example, an Accumulated Interior Gateway Protocol (“AIGP”) attribute, a non-transitive BGP attribute, which includes a link metric assigned to a link within a first IGP domain, is scaled to conform to a metric scale of the second IGP domain. The AIGP attribute may also add link metric assigned to a link within the second IGP domain and may add static metrics of non-IGP links connecting the IGP domains. An IGP domain may set its IGP to the scaled AIGP attribute such that the link metric may include a uniformly scaled end-to-end view of link metrics across the IGP domains. Additionally, a sham-link is assigned a metric value in accordance with the scaling techniques.

Abstract: In general, techniques are described for reducing or otherwise preventing micro-loops in network using Source Packet Routing in Networking (SPRING). In some examples, a method includes detecting a failure of a communication by a network device that implements a Source Packet Routing in Networking (SPRING) protocol to forward network packets using node labels according to an initial network topology. Responsive to detecting the failure of the communication link, the network device may apply, for a defined time duration, one or more adjacency labels to network packets to define a set of one-hop tunnels corresponding to a backup sub-path that circumvents the failed communication link. Upon expiration of the defined time duration, the network device may forward, according to a new network topology that is not based on applying the one or more adjacency labels that define the set of one-hop tunnels, network packets destined for the destination network device.

Abstract: In general, techniques are generally described for reducing or preventing transient black-holing of network traffic in an overlay network. A method includes executing, by a network device included in a link state domain, an Interior Gateway Protocol (IGP) to exchange link-state messages with at least one remote network device in the link-state domain; generating, by the network device, an IGP link-state message that includes link overload information to overload a link in the link-state domain that couples the network device to the remote network device; and sending, by the network device and to the at least one other network device, the IGP link-state message that includes the link overload information to direct the remote network device to stop sending network traffic to the network device using the overloaded link.

Abstract: A routing device coupled to a remote routing device via a link on which a flood reduction technique is used, such as a demand circuit, is configured to store an indication of a link state of the remote routing device and a first sequence number associated with the link state, receive an indication that the remote routing device is performing a graceful restart, and then receive data indicative of a new link state of the remote routing device and a second sequence number. The routing device determines whether the new link state is different than the stored indication of the link state, and if not, avoids requesting the current link state from the remote routing device. In this manner, the routing device may reduce link-state protocol traffic within an autonomous system including the routing device and the remote routing device.

Abstract: In general, techniques are described for managing routing information in a hub-and-spoke network in a manner that reduces flooding of link information. A hub router of the hub-and-spoke network including a memory and a processor may perform the techniques. The memory may be configured to store a representation of a topology of the hub-and-spoke network. The processor may be configured to utilize a separate instance of a multi-instance version of a link state protocol to communicate with each of a plurality of spoke routers of the hub-and-spoke network. Each separate instance of the multi-instance version of the link state protocol may include the hub router and a different one of the plurality of spoke routers. The processor may process link state advertisements from the separate instances of the multi-instance version of the link state protocol to maintain the representation of the topology of the hub-and-spoke network.

Abstract: In general, techniques are generally described for reducing or preventing transient black-holing of network traffic in an overlay network. A first customer edge (CE) network device positioned in a first customer network may be configured to perform the techniques. The first CE network device may comprise a control unit configured to execute an instance of a network protocol to detect faults between the first CE network device and a second CE network device positioned in a second customer network. The first CE network device may also comprise an interface configured to transmit a message to the second CE network device via the instance of the network protocol signaling that a provider edge (PE) network device is going to become nonoperational. The PE network device may be positioned in an intermediate network providing interconnectivity between the first customer network and the second customer network.

Abstract: An example method includes selecting, by a network device, a remote LFA next hop as an alternate next hop for forwarding network traffic from the network device to a destination, wherein the selected remote LFA next hop provides node protection to a primary next hop node on the shortest path from the network device to the destination. The method includes, for each candidate remote LFA next hop, performing a forward shortest path first (SPF) computation having the respective candidate remote LFA next hop as a root to compute a path segment between the respective candidate remote LFA next hop and the destination, wherein each of the candidate remote LFA next hops is the egress of a respective potential repair tunnel between the network device and candidate remote LFA next hop, and selecting the remote LFA next hop based at least in part on the computed path segments.

Abstract: In one example, a network device determines a set of candidate loop-free alternate (LFA) next hops for forwarding network traffic from the network device to a multi-homed network by taking into account a first cost associated with a second path from a first border router to the multi-homed network and a second cost associated with a second border router to the multi-homed network, wherein the multi-homed network is external to an interior routing domain in which the network device is located. The network device selects an LFA next hop from the set of candidate LFA next hops, to be stored as an alternate next hop for forwarding network traffic to the multi-homed network, and updates forwarding information stored by the network device to install the selected LFA next hop as the alternate next hop for forwarding network traffic from the network device to the multi-horned network.

Abstract: A routing device coupled to a remote routing device via a link on which a flood reduction technique is used, such as a demand circuit, is configured to store an indication of a link state of the remote routing device and a first sequence number associated with the link state, receive an indication that the remote routing device is performing a graceful restart, and then receive data indicative of a new link state of the remote routing device and a second sequence number. The routing device determines whether the new link state is different than the stored indication of the link state, and if not, avoids requesting the current link state from the remote routing device. In this manner, the routing device may reduce link-state protocol traffic within an autonomous system including the routing device and the remote routing device.

Abstract: In general, techniques are described for managing routing information in a hub-and-spoke network in a manner that reduces flooding of link information. A hub router of the hub-and-spoke network including a memory and a processor may perform the techniques. The memory may be configured to store a representation of a topology of the hub-and-spoke network. The processor may be configured to utilize a separate instance of a multi-instance version of a link state protocol to communicate with each of a plurality of spoke routers of the hub-and-spoke network. Each separate instance of the multi-instance version of the link state protocol may include the hub router and a different one of the plurality of spoke routers. The processor may process link state advertisements from the separate instances of the multi-instance version of the link state protocol to maintain the representation of the topology of the hub-and-spoke network.

Abstract: In general, techniques are generally described for reducing or preventing transient black-holing of network traffic in an overlay network. A method includes executing, by a network device included in a link state domain, an Interior Gateway Protocol (IGP) to exchange link-state messages with at least one remote network device in the link-state domain; generating, by the network device, an IGP link-state message that includes link overload information to overload a link in the link-state domain that couples the network device to the remote network device; and sending, by the network device and to the at least one other network device, the IGP link-state message that includes the link overload information to direct the remote network device to stop sending network traffic to the network device using the overloaded link.

Abstract: In one example, a network device determines a set of candidate loop-free alternate (LFA) next hops for forwarding network traffic from the network device to a multi-homed network by taking into account a first cost associated with a second path from a first border router to the multi-homed network and a second cost associated with a second border router to the multi-homed network, wherein the multi-homed network is external to an interior routing domain in which the network device is located. The network device selects an LFA next hop from the set of candidate LFA next hops, to be stored as an alternate next hop for forwarding network traffic to the multi-homed network, and updates forwarding information stored by the network device to install the selected LFA next hop as the alternate next hop for forwarding network traffic from the network device to the multi-horned network.