Abstract: A network system for a data center is described in which a switch fabric provides full mesh interconnectivity such that any servers may communicate packet data to any other of the servers using any of a number of parallel data paths. Moreover, according to the techniques described herein, edge-positioned access nodes, optical permutation devices and core switches of the switch fabric may be configured and arranged in a way such that the parallel data paths provide single L2/L3 hop, full mesh interconnections between any pairwise combination of the access nodes, even in massive data centers having tens of thousands of servers.

Abstract: A fabric control protocol is described for use within a data center in which a switch fabric provides full mesh interconnectivity such that any of the servers may communicate packet data for a given packet flow to any other of the servers using any of a number of parallel data paths within the data center switch fabric. The fabric control protocol enables spraying of individual packets for a given packet flow across some or all of the multiple parallel data paths in the data center switch fabric and, optionally, reordering of the packets for delivery to the destination. The fabric control protocol may provide end-to-end bandwidth scaling and flow fairness within a single tunnel based on endpoint-controlled requests and grants for flows. In some examples, the fabric control protocol packet structure is carried over an underlying protocol, such as the User Datagram Protocol (UDP).

Abstract: A network system for a data center is described in which a switch fabric provides full mesh interconnectivity such that any servers may communicate packet data to any other of the servers using any of a number of parallel data paths. Moreover, according to the techniques described herein, edge-positioned access nodes, permutation devices and core switches of the switch fabric may be configured and arranged in a way such that the parallel data paths provide single L2/L3 hop, full mesh interconnections between any pairwise combination of the access nodes, even in massive data centers having tens of thousands of servers. The access nodes may be arranged within access node groups, and permutation devices may be used within the access node groups to spray packets across the access node groups prior to injection within the switch fabric, thereby increasing the fanout and scalability of the network system.

Abstract: Systems, methods, and non-transitory computer-readable storage media for stage upgrades in a network. The system generates graph-data structured based representations of devices in the network, wherein respective attributes of the representations is selected based on respective services provided by the devices to tenants in the network and identities of respective tenants serviced by the devices. Next, the system generates a graph showing a distribution of the devices in the network according to the representations, wherein the representations are interconnected in the graph based on service roles of associated devices with respect to tenants in the network and other devices associated with the tenants. The system then schedules an upgrade of devices based on the graph, the upgrade being scheduled in stages, each stage including devices selected for upgrade in that stage, wherein the devices for each stage are selected by identifying devices having respective representations assigned to that specific stage.

Abstract: Systems, methods, and non-transitory computer-readable storage media for recovering from a partial failure of a virtual port chain (vPC) domain. The first and second vPC peers may be paired to create a vPC having a virtual address. An endpoint host may communicate with a network via the virtual port channel. The system may detect that the first virtual port channel peer is down. During or after the first vPC reboots, the reachability cost for the first vPC with regards to the virtual address can be set to an inflated value. The first vPC peer may also delay its bring up time while it synchronizes its vPC state information with the second vPC peer. The second vPC can continue to advertise the association between the endpoint host and the virtual address. Upon completion of the synchronization, the first vPC peer may bring up the link and restore the reachability cost.

Abstract: A packet is generated at a first network connected device for transmission to a destination network device through a network comprising a plurality of pods. At least two of the plurality of pods are within separate management domains, and generating the packet comprises generating the packet with a first identifier and a second identifier. The first identifier indicates a pod of the plurality of pods in which the destination network connected device is located, and the second identifier indicates an identity of the destination network connected device within the pod of the plurality of pods. The packet is transmitted from the first network connected device to the destination network connected device.

Abstract: A fabric control protocol is described for use within a data center in which a switch fabric provides full mesh interconnectivity such that any of the servers may communicate packet data for a given packet flow to any other of the servers using any of a number of parallel data paths within the data center switch fabric. The fabric control protocol enables spraying of individual packets for a given packet flow across some or all of the multiple parallel data paths in the data center switch fabric and, optionally, reordering of the packets for delivery to the destination. The fabric control protocol may provide end-to-end bandwidth scaling and flow fairness within a single tunnel based on endpoint-controlled requests and grants for flows. In some examples, the fabric control protocol packet structure is carried over an underlying protocol, such as the User Datagram Protocol (UDP).

Abstract: Techniques for detecting path failures and reducing packet loss as a result of such failures are described for use within a data center or other environment. For example, a source and/or destination access node may create and/or maintain information about health and/or connectivity for a plurality of ports or paths between the source and destination device and core switches. The source access node may spray packets over a number of paths between the source access node and the destination access node. The source access node may use the information about connectivity for the paths between the source or destination access nodes and the core switches to limit the paths over which packets are sprayed. The source access node may spray packets over paths between the source access node and the destination access node that are identified as healthy, while avoiding paths that have been identified as failed.

Abstract: Network access node virtual fabrics configured dynamically over an underlay network are described. A centralized controller, such as a software-defined networking (SDN) controller, of a packet switched network is configured to establish one or more virtual fabrics as overlay networks on top of the physical underlay network of the packet switched network. For example, the SDN controller may define multiple sets of two of more access nodes connected to the packet switched network, and the access nodes of a given one of the sets may use a new data transmission protocol, referred to generally herein as a fabric control protocol (FCP), to dynamically setup tunnels as a virtual fabric over the packet switched network. The FCP tunnels may include all or a subset of the parallel data paths through the packet switched network between the access nodes for a given virtual fabric.

Abstract: Systems, methods and transitory computer-readable storage media for constructing a loop free multicast tree. The methods include observing a network topology transition affecting a first path from the particular node to a root node, calculating a second path from the particular node to the root node and sending a message to an upstream node requesting that the upstream node be a root port in the calculated second path. If the upstream node agrees to be the root port in the calculated second path, the method further includes creating a new FTAG tree topology view that includes the upstream node as the root port in the second path.

Abstract: Systems, methods, and non-transitory computer-readable storage media for recovering from a partial failure of a virtual port chain (vPC) domain. The first and second vPC peers may be paired to create a vPC having a virtual address. An endpoint host may communicate with a network via the virtual port channel. The system may detect that the first virtual port channel peer is down. During or after the first vPC reboots, the reachability cost for the first vPC with regards to the virtual address can be set to an inflated value. The first vPC peer may also delay its bring up time while it synchronizes its vPC state information with the second vPC peer. The second vPC can continue to advertise the association between the endpoint host and the virtual address. Upon completion of the synchronization, the first vPC peer may bring up the link and restore the reachability cost.

Abstract: A network system for a data center is described in which a switch fabric provides full mesh interconnectivity such that any servers may communicate packet data to any other of the servers using any of a number of parallel data paths. Moreover, according to the techniques described herein, edge-positioned access nodes, permutation devices and core switches of the switch fabric may be configured and arranged in a way such that the parallel data paths provide single L2/L3 hop, full mesh interconnections between any pairwise combination of the access nodes, even in massive data centers having tens of thousands of servers. The access nodes may be arranged within access node groups, and permutation devices may be used within the access node groups to spray packets across the access node groups prior to injection within the switch fabric, thereby increasing the fanout and scalability of the network system.

Abstract: A network system for a data center is described in which a switch fabric provides full mesh interconnectivity such that any servers may communicate packet data to any other of the servers using any of a number of parallel data paths. Moreover, according to the techniques described herein, edge-positioned access nodes, optical permutation devices and core switches of the switch fabric may be configured and arranged in a way such that the parallel data paths provide single L2/L3 hop, full mesh interconnections between any pairwise combination of the access nodes, even in massive data centers having tens of thousands of servers.

Abstract: A determination is made at a network connected device that a network policy is to be verified. The network policy is applied to network packets sent to an endpoint within a network, and the application of the policy to network traffic can result in at least two outcomes. Another determination is made at the network connected device that a switch is provisionable to host the endpoint. The network connected device provisions a simulated endpoint version of the endpoint at the switch to host the policy. At least one packet is sent to the simulated endpoint via the network connected device for each of the at least two outcomes of the policy. At least one response is received by the network connected device from the simulated endpoint indicating how the policy was applied to each of the packets.

Abstract: Exemplary embodiments identify all viable paths in an ECMP/WCMP enabled network without running traceroute multiple times. Devices in a network may be configured to send a packet including a pre-determined option to a Software-Defined Network Controller (SDNC) upon receipt. If a destination of the packet is within the SDNC-controlled domain, SDNC identifies all viable ECMP/WCMP paths using routing information of the network. If the destination of the packet is outside the SDNC-controlled domain, SDNC identifies at least one egress switch of the SDNC-controlled domain. SDNC may identify internal paths connecting the source of the packet to the at least one egress switch. SDNC may also identify external paths connecting the at least one egress switch to the destination of the packet. SDNC may construct the viable ECMP/WCMP paths by connecting the internal paths to the external paths.

Abstract: Systems, methods and non-transitory computer-readable storage media for determining interconnectivity in dense networks. The method includes generating, at a first network device in a data network, a traceroute packet that includes source and destination address information. The packet is encapsulated in an outer packet, and the encapsulated packet is forwarded to a second network device and to one or more intermediate network devices in the data network. A response packet is received from the second network device and each of the intermediate network devices that received the traceroute packet. The first network device determines, based on the responsive packets, an end-to-end path taken by the traceroute packet through the data network.

Abstract: An inner packet configured with a multicast address and configured to perform a traceroute operation through a network is encapsulated to form an encapsulated packet. The encapsulated packet is sent into a network, the encapsulated packet being forwarded along a multicast tree of the network for the multicast address. A plurality of responses are received from a plurality of network nodes comprising the multicast tree, wherein each response comprises an indication of a node of the plurality of nodes that sends the response and an indication of a node from which the node sending the response received the encapsulated packet.

Abstract: Systems, methods and transitory computer-readable storage media for constructing a loop free multicast tree. The methods include observing a network topology transition affecting a first path from the particular node to a root node, calculating a second path from the particular node to the root node and sending a message to an upstream node requesting that the upstream node be a root port in the calculated second path. If the upstream node agrees to be the root port in the calculated second path, the method further includes creating a new FTAG tree topology view that includes the upstream node as the root port in the second path.

Abstract: Exemplary embodiments provide a client in a network sending data packets to a server using multiple paths. The client may check if the server can receive packets sent at the backup ports of the client by including an option in a first packet sent to the server. The option included in the packet may provide a list of available backup ports that may be used by the client to communicate with the server. If the server supports the option, the server includes the option in an acknowledgment packet sent back to the client. The client and the server may create a mapping from the client's backup ports to the client's primary port. Thus, when the server receives a packet sent at a backup port, the server treats the packet as if the packet was sent at the primary port of the client.

Abstract: A packet is generated at a first network connected device for transmission to a destination network device through a network comprising a plurality of pods. At least two of the plurality of pods are within separate management domains, and generating the packet comprises generating the packet with a first identifier and a second identifier. The first identifier indicates a pod of the plurality of pods in which the destination network connected device is located, and the second identifier indicates an identity of the destination network connected device within the pod of the plurality of pods. The packet is transmitted from the first network connected device to the destination network connected device.