Systems, methods, and software are disclosed herein for routing in-bound communications to an infrastructure service. In an implementation, an infrastructure service receives a request from an end point for content associated with an origin. The service sends a connection request to the origin from an initial network address. After detecting a failure of the origin to respond to the connection request, the service sends multiple connection requests to the origin from different network addresses. Upon receiving one or more replies to the connection requests, the service identifies which reply was received first and a network address to which the reply was sent. The service proceeds to establish a connection with the origin using the identified network address and obtains the content from the origin over the connection. The infrastructure service may then send the content to the end point.

A network device includes processing circuitry and multiple ports. The multiple ports are configured to connect to a communication network. The processing circuitry is configured to select a first port among the multiple ports to serve as an egress port for a packet, and to forward the packet to the first port, irrespective of whether or not the first port is usable as the egress port. The processing circuitry is further configured to, when the first port is usable as the egress port, transmit the packet to the communication network via the first port, and when the first port is unusable as the egress port, forward the packet from the first port to a second port among the multiple ports and transmit the packet to the communication network via the second port.

In one embodiment, a supervisory service for a software-defined wide area network (SD-WAN) obtains telemetry data from one or more edge devices in the SD-WAN. The service trains, using the telemetry data as training data, a machine learning-based model to predict tunnel failures in the SD-WAN. The service receives feedback from the one or more edge devices regarding failure predictions made by the trained machine learning-based model. The service retrains the machine learning-based model, based on the received feedback.

Systems, methods, and devices for improved routing operations in a network computing environment. A system includes a virtual customer edge router and a host routed overlay comprising a plurality of host virtual machines. The system includes a routed uplink from the virtual customer edge router to one or more of the plurality of leaf nodes. The system is such that the virtual customer edge router is configured to provide localized integrated routing and bridging (IRB) service for the plurality of host virtual machines of the host routed overlay.

A method by a network device functioning as a provider edge (PE) in an ethernet virtual private network (EVPN) to prevent transient loops between multi-home peer PEs. The method includes advertising a first EVPN label to one or more PEs that are multi-home peer PEs of the PE, advertising a second EVPN label to one or more PEs that are not multi-home peer PEs of the PE, receiving first traffic for a CE that is encapsulated with the first EVPN label as opposed to the second EVPN label, and discarding the first traffic in response to determining that a link between the PE and the CE is not operational and the first traffic for the CE is encapsulated with the first EVPN label.

Method and computing devices for enforcing packet order based on packet marking. Upon occurrence of a link failure, a first device reallocates traffic initially forwarded through the failed link to an alternative link and marks the reallocated traffic with a first flag. Upon recovery of the failed link, the reallocated traffic is forwarded again through the recovered link and marked with a second flag different from the first flag. A second device calculates a reference inter-packet time for received traffic marked with the first flag. For received traffic marked with the second flag, the second device calculates a current inter-packet time. The current inter-packet time is compared with the reference inter-packet time, to determine if the traffic marked with the second flag shall be forwarded immediately or if the forwarding shall be delayed.

Systems and methods associated with a node in a Segment Routing network include, responsive to what services are support at a node in a Segment Routing network, creating a bitmap to represent the plurality of services supported at the node; and transmitting an advertisement with the bitmap such that the advertisement is a single advertisement of multiple services. This approach can reduce the advertisement of rout updates by orders of magnitude.

A system is disclosed, comprising: a centralized routing node configured to: identify a set of congested links based on the link utilization statistics, each congested link having at least one traffic flow that may be active, each traffic flow having at least one traffic source and a path set comprising a set of nodes and links that may be used by the traffic flow as packets travel from the at least one traffic source to one or more destinations; identify a set of non-congested links based on the link utilization statistics, each non-congested link sharing at least one traffic source with a traffic flow of a congested link in the set of congested links; identify a path fork in a path set between a source and a destination of a particular traffic flow associated with a particular congested link in the set of congested links; and compute a new utilization level for the particular congested link that would result from moving the particular traffic flow from the particular congested link to a particular non-congested link in the set of non-congested links.

A method for assessing path quality between a source node and a destination node, the method comprising, at a node: determining a first parameter and a first path classification of a first path having at least one link; determining a second parameter and a second path classification of a second path having at least one link; determining the quality of the first path based on the first parameter and the first path classification; determining the quality of the second path based on the second parameter and the second path classification; and comparing the quality of the first path and the quality of the second path thereby to determine the best quality path.

A packet loss processing method and a network device are provided. The method includes: A first node obtains a first forwarding label of a first packet, where the first packet is a discarded packet. The first node determines, based on the first forwarding label, that the first node does not have a LSP corresponding to the first forwarding label. The first node sends a first message to a second node, where the first message includes the first forwarding label, and the first message is used to indicate that the first node does not have the LSP corresponding to the first forwarding label. The second node may be, for example, a peer node of the first node. The first node sends the message to the peer node, to indicate that the first node does not have the LSP corresponding to the forwarding label.