Cisco Technology, Inc.

Jianda Liu

Xiaorong Wang

Hongbo Xia

Pritam Baruah

Yicheng Liu

Ruchir Jain

Olivier Patrick Jean Pelerin

Frederic René Philippe Detienne

Jiayi You

Yunpeng Zhang

Shen Yong Qing

Yu Zhang

Lianxiang Wang

Gaofeng Tao

Bin Shi

Feng Ping Cai

Zhiyong Fang

Xiao-Rong Wang

Danmu Wu

Wei Zhou

Xiangyu Ji

Techniques and architecture are described for tracking flows of packets in a network using a packet color marking scheme for obtaining network end-to-end traffic flow metrics in the network. In particular, in configurations, a user may be prompted to specify a site and a virtual private network (VPN) at which to start a trace for packet flows using a coloring marking scheme within the network. Given the VPN and the site, it is possible to monitor interested packet flows and apply metadata, e.g., colors, on flow packets. Remote wide area network (e.g., software defined WAN (SD-WAN)) routers receiving the packets with metadata may automatically trace the same flow and apply the same metadata to a next hop router. Hence, an end-to-end network path may be discovered.

Network end-to-end traffic flow statistics from arbitrary network location

This disclosure describes techniques and mechanisms for intelligently sampling packet flows within a network. The techniques enable the sampling of a limited set of packet flows that show greatest amount of information about the network from the packet flows in order to provide the greatest insight on application performance, network packet, and critical events within the network. Additionally, the techniques provide configurable parameters, such that the techniques are customizable for each user's network.

Packet flow sampling in network monitoring

According to an embodiment, a node comprises one or more processors operable to execute instructions to cause the node to perform operations that comprise receiving a packet from a first node associated with an SD-WAN domain. The packet comprises a header indicating a TLOC associated with a second node to send the packet, the second node associated with an SR domain. The operations comprise determining that the TLOC corresponds to a virtual TLOC used in the SD-WAN domain to identify the second node that is in the SR domain and, in response, determining a second node identifier used in the SR domain to identify the second node. The operations further comprise preparing the packet to be communicated via the SR domain. Preparing the packet comprises including the second node identifier in the packet. The operations further comprise sending the packet comprising the second node identifier to the second node.

Inter-working of a software-defined wide-area network (SD-WAN) domain and a segment routing (SR) domain

Techniques and architecture are described grouping various sources of traffic within a network into grouping fields and assigning each combination of grouping field values an aggregate identification (ID). A first hop edge router may receive a packet and search a mapping table for a corresponding aggregate ID for the combination of grouping field values within the mapping table. If not found, the first hop edge router may assign a corresponding aggregate ID for the combination of grouping field values and store the new aggregate ID for the combination of grouping field values in the mapping table. The first hop edge router may forward the packet on through the network with the aggregate ID embedded in metadata. Routers within the network may measure and aggregate flow metrics of the packet within the network based on the aggregate ID and provide the measurements to the network controller.

Dynamic aggregate ID based flow metrics aggregation

In one embodiment, a method includes receiving, by a network orchestrator, trace parameters from a user device. The method also includes determining, by the network orchestrator, to initiate a network path trace for the application, generating, by the network orchestrator, a filter policy for the network path trace using the trace parameters, and allocating, by the network orchestrator, a trace identification to the network path trace. The method also includes initiating, by the network orchestrator, the network path trace within a network by communicating the filter policy and the trace identification to a first node of the network and receiving, by the network orchestrator, network path trace data from a plurality of nodes of the network. The method further includes generating, by the network orchestrator, a trace report for the application using the network path trace data.

Systems and methods for determining a network path trace

A method is provided in one embodiment and includes receiving at a network element an encapsulated packet and determining whether both an ECMP/LAG Existing ('ele') flag and an Entropy Label Capability ('elc') flag are set for an egress node of the packet in a Label Distribution Protocol (&#x201c;LDP&#x201d;) database of the network element. If both the ele and elc flags are set for the egress node of the packet in the LDP database, the method further includes determining whether the network element is an ingress node for the packet and, if the network element is the ingress node for the packet, pushing an Entropy Label (&#x201c;EL&#x201d;) and an Entropy Label Indicator (&#x201c;ELI&#x201d;) onto an MPLS stack of the packet.

Techniques for decreasing multiprotocol label switching entropy label overhead

In an example, a server architecture is described for a dynamic cascaded node chain providing a resource cluster. The cascaded node chain may include one or more resource instances provisioned as a head node, zero or more middle nodes, and a tail node. Each node may include a discrete number of available resource entries in a flow table. As traffic enters the head node, each node attempts to match the traffic to an entry in its flow table. If no match is found, the packet is downlinked to the next node in the chain. If the packet reaches the tail node without a match, it is punted to the controller. The controller may then provision a matching entry if an entry is available. If not, the controller may spawn a new resource instance. When the full capacity of the cluster is reached, non-matching entries may be dropped.

Dynamic cascaded clustering for dynamic VNF

Growing community of inventors

Shanghai, China

Jianda Liu