Abstract: Methods and system for directing traffic flows to a fast data path or a slow data path are disclosed. Parsers can produce packet header vectors (PHVs) for use in match-action units. The PHVs are also used to generate feature vectors for the traffic flows. A flow training engine produces a classification model. Feature vectors input to the classification model result in output predictions predicting if a traffic flow will be long lived or short lived. The classification models are used by network appliances to install traffic flows into fast data paths or the slow data paths based on the predictions.

Abstract: A network appliance or smart switch can include service devices as well as a switching device such as those used in high-speed switches having limited processing ability and are stateless with respect to sessions. Service devices can provide stateful and complex processing. A first exposed port of a switching device can receive network packets and can determine which network packets the service devices are to process to produce processed network packets. A network packet can be sent to a service device in a redirected packet. A processed network packet can be received from a service device in a reinjected packet that is used to recover a port identifier of the first exposed port. The port identifier can be used to determine a network destination of the processed network packet. The processed network packet can be sent from a second exposed port of the switching device toward the network destination.

Abstract: Methods and system for directing traffic flows to a fast data path or a slow data path are disclosed. Parsers can produce packet header vectors (PHVs) for use in match-action units. The PHVs are also used to generate feature vectors for the traffic flows. A flow training engine produces a classification model. Feature vectors input to the classification model result in output predictions predicting if a traffic flow will be long lived or short lived. The classification models are used by network appliances to install traffic flows into fast data paths or the slow data paths based on the predictions.

Abstract: Methods and system for directing traffic flows to a fast data path or a slow data path are disclosed. Parsers can produce packet header vectors (PHVs) for use in match-action units. The PHVs are also used to generate feature vectors for the traffic flows. A flow training engine produces a classification model. Feature vectors input to the classification model result in output predictions predicting if a traffic flow will be long lived or short lived. The classification models are used by network appliances to install traffic flows into fast data paths or the slow data paths based on the predictions.

Abstract: Methods and system for directing traffic flows to a fast data path or a slow data path are disclosed. Parsers can produce packet header vectors (PHVs) for use in match-action units. The PHVs are also used to generate feature vectors for the traffic flows. A flow training engine produces a classification model. Feature vectors input to the classification model result in output predictions predicting if a traffic flow will be long lived or short lived. The classification models are used by network appliances to install traffic flows into fast data paths or the slow data paths based on the predictions.

Abstract: An example method for to multicast traffic distribution in a multi-pod network environment is provided and includes provisioning a block of multicast group addresses for broadcast, unknown unicast and multicast (BUM) traffic distribution between pods in the multi-pod network, calculating a hash corresponding to a bridge domain (BD) extending across a plurality of pods in the multi-pod network, the hash being identically calculated at each one of the plurality of pod, indexing with the hash into the block of multicast group addresses designated for inter-pod BUM traffic to derive a global multicast group identical for the broadcast domain across the plurality of pods, and associating a local multicast group at the translator with the derived global multicast group.

Abstract: An example method for to multicast traffic distribution in a multi-pod network environment is provided and includes provisioning a block of multicast group addresses for broadcast, unknown unicast and multicast (BUM) traffic distribution between pods in the multi-pod network, calculating a hash corresponding to a bridge domain (BD) extending across a plurality of pods in the multi-pod network, the hash being identically calculated at each one of the plurality of pod, indexing with the hash into the block of multicast group addresses designated for inter-pod BUM traffic to derive a global multicast group identical for the broadcast domain across the plurality of pods, and associating a local multicast group at the translator with the derived global multicast group.

Abstract: Presented herein are embodiments for tracing paths of packet flows in a data center fabric network. Filters are configured on nodes (e.g., switches) in the data center fabric network for a particular packet flow. Numerous such filters can be configured on each of the switches, each filter for a different packet flow. When a filter detects a match, it sends a log of such occurrence to a network controller. The network controller uses log data sent from nodes as well as knowledge of the network topology (updated as changes occur in the network) to determine the path for a particular packet flow in the data center fabric network. This technique works inline on the actual packet flow and does not need additional debug packets to be injected. This technique can also quickly point out the problem node in case of traffic drop.