Abstract: In one embodiment, a service receives telemetry data indicative of a plurality of performance metrics captured in a network. The service jointly trains, using the received telemetry data, a compression model and an inference model, the compression model being a first machine learning model trained to convert the telemetry data into a compressed representation of the telemetry data and the inference model being a second machine learning model trained to take the compressed representation of the telemetry data as input and apply a classification label to it. The service deploys the compression model to the network. The service receives compressed telemetry data generated by the compression model deployed to the network. The service uses the inference model to classify the compressed telemetry data generated by the compression model deployed to the network.

Abstract: This disclosure describes techniques for applying a policy proximate to a source of data traffic in a network. The techniques include indicating to a destination edge node that a policy relevant to the data traffic has not been applied at a source edge node. The destination edge node may send the policy to the source edge node. The source edge node may apply the policy to a subsequent packet of the data traffic. Application of the policy proximate to the source of the data traffic may conserve network resources and improve performance of the network.

Abstract: This disclosure describes techniques for providing a distributed scalable architecture for Network Address Translation (NAT) systems with high availability and mitigations for flow breakage during failover events. The NAT servers may include functionality to serve as fast-path servers and/or slow-path servers. A fast-path server may include a NAT worker that includes a cache of NAT mappings to perform stateful network address translation and to forward packets with minimal latency. A slow-path server may include a mapping server that creates new NAT mappings, depreciates old ones, and answers NAT worker state requests. The NAT system may use virtual mapping servers (VMSs) running on primary physical servers with state duplicated VMSs on different physical failover servers.

Abstract: Systems, methods, and computer-readable storage media are provided for managing connected data transfer sessions in a computing network. A controller included in the computing network can monitor connected data transfer sessions to determine whether a predetermined threshold has been met or exceeded and, if so, terminate at least one connected data transfer session in the computing network. The threshold can include a threshold number of connected data communication sessions and/or a threshold amount of resources utilized by the connected data communication sessions. The controller can terminate connected data transfer sessions until the total number of connected data communication sessions and/or threshold amount of resources falls below the threshold.

Abstract: In one embodiment, a device deploys a first machine learning model to an inference location in a network. The first machine learning model is used at the inference location to make inferences about the network. The device receives, from the inference location, an indication that the first machine learning model is exhibiting poor performance. The device identifies a corrective measure for the poor performance that minimizes resource consumption by a model training pipeline of the device. The device deploys, based on the corrective measure, a second machine learning model to the inference location. The second machine learning model is used in lieu of the first machine learning model to make the inferences about the network.

Abstract: Techniques for connectivity issue remediation are provided. A first link trace message is automatically transmitted from a source end point to a destination end point. A first topology graph is generated for a network based on the first link trace message. Presence of a loop is detected in the network. A second link trace message is then transmitted from the source end point to the destination end point, and a second topology graph is automatically generated for the network based on the second link trace message. An edge in the network that caused the loop is identified based on comparing the first and second topology graphs.

Abstract: In one embodiment, a Segment Routing network node provides efficiencies in processing and communicating Internet Protocol packets in a network. This Segment Routing node typically advertises (e.g., using Border Gateway Protocol) its Segment Routing processing capabilities, such as Penultimate Segment Pop (PSP) and/or Ultimate Segment Pop (USP) of a Segment Routing Header (including in the context of a packet that has multiple Segment Routing Headers). Subsequently, an Internet Protocol Segment Routing packet having multiple Segment Routing Headers is received. The packet is processed according to a Segment Routing function, with is processing including removing a first one of the Segment Routing Headers and forwarding the resultant Segment Routing packet. The value of the Segments Left field in the first Segment Routing Header identifies to perform PSP when the value is one, to perform USP when the value is zero, or to perform other processing.

Abstract: Techniques for providing Hybrid information-centric networking (hICN) via a proxy application is described. A hICN proxy application provides hICN to legacy applications by diverting network traffic of a plurality of network traffic types to the hICN proxy application and storing network traffic information for the network traffic in a connection table. The hICN proxy application also translates the diverted network traffic to a hICN network traffic protocol and selects a forwarding strategy for the translated network traffic in order to send the hICN traffic over various non-hICN network protocol types. The hICN proxy application also transmits the translated traffic to a server proxy application using the selected forwarding strategy.

Abstract: A method includes, at a media bridge configured to distribute a plurality of media streams among a plurality of client devices connected to the media bridge over a network during a real-time communication (RTC) instance, receiving a plurality of quality of experience (QoE) preferences from the plurality of client devices via the media bridge, the plurality of QoE preferences being transmitted as a real-time transport protocol (RTP) control protocol (RTCP) extension header of a transmitted data packet. The method also includes receiving a plurality of QoE metrics at the media bridge, and in response to a determination that a degradation in network conditions of the network has occurred, downgrading at least one of the plurality of media streams based on the plurality of QoE preferences.

Abstract: Techniques and mechanisms for a control plane approach for dense topologies that focusses on discovering shared ECMP groups in the control plane independent of per-prefix learning and then learning prefixes via these shared ECMP groups instead of learning prefixes via one next-hop at a time. In dense topologies, this approach helps minimize BGP path scale, corresponding signaling and enables control plane scaling that is an order of magnitude higher than a traditional eBGP control plane. During link and node topology changes, the described control plane approach enables control plane signaling that is prefix independent and an order of magnitude lower. A control plane approach to path-list sharing and prefix independent signaling on link and node topology changes enables prefix independent convergence (PIC) in scenarios that would not be possible otherwise with traditional FIB driven path-list sharing and PIC.

Abstract: Techniques and apparatus for managing congestion in a wireless network are provided. One technique includes receiving one or more buffer status reports (BSRs) from one or more client stations. Each BSR indicates an amount of traffic in a transmit queue of the client station. An allocation of resource units (RUs) for the one or more client stations is determined, based on at least a type of traffic in each transmit queue of the one or more client stations, upon a determination that there is congestion on an uplink wireless channel shared by the one or more client stations. A frame that includes an indication of the allocation of RUs for the one or more client stations is generated. The frame is transmitted to the one or more client stations.

Abstract: According to one or more embodiments of the disclosure, a first tunnel router may receive a reservation request to establish a deterministic path between a first node and a second node. The first tunnel router may determine, based on the reservation request, a destination address of the second node. The first tunnel router may identify, based on the destination address of the second node, a second tunnel router associated with the second node. The first tunnel router may encapsulate a deterministic packet sent by the first towards the second node into a tunnel packet, wherein a multicast address in a header of the tunnel packet is set to the destination address of the second node. The first tunnel router can forward the tunnel packet along the deterministic path. The multicast address in the header of the tunnel packet causes nodes to send the tunnel packet according to the deterministic path.

Abstract: Embodiments herein describe coupling traditional fan and shaper control along with aggregated knowledge of the temperature history of a hardware device to optimally manage the temperature of the hardware device to preserve its expected life while also providing the lower power, best performing solution possible. In one embodiment, a cooling application manages the expected life by trading off performance and power versus temperature to achieve a desired (or accepted) lifetime. In one embodiment, the cooling application calculates a historical temperature value for the hardware device which is then used to determine the expected life of the hardware device.

Abstract: Various implementations disclosed herein enable malleable routing for data packets. For example, in various implementations, a method of routing a type of data packets is performed by a device. In some implementations, the device includes a non-transitory memory and one or more processors coupled with the non-transitory memory. In some implementations, the method includes determining a routing criterion to transmit a set of data packets across a network. In some implementations, the method includes identifying network nodes and communication links in the network that satisfy the routing criterion. In some implementations, the method includes determining a route for the set of data packets through the network nodes and the communication links that satisfy the routing criterion. In some implementations, the method includes configuring the network nodes that are on the route with configuration information that allows the set of data packets to propagate along the route.

Abstract: An identity provider (IdP) service interoperates with a Virtual Private Network (VPN) client. The IdP service receives a login request originating from the VPN client to establish a VPN tunnel between the VPN client and a VPN host, the login request indicating a user of the VPN client. The IdP service provides a response to the login request. The response includes at least both first information including an indication that the user of the VPN client is an authorized user and second information including an indication of a VPN policy for the VPN tunnel, the VPN policy including a VPN client policy to be utilized during the VPN tunnel by the VPN client and a VPN host policy to be utilized during the VPN tunnel by the VPN host.

Abstract: Protocol independent signal slotting and scheduling is provided by receiving a frame including a header and a payload for transmission; in response to determining that the frame matches a rule identifying the frame as part of a control loop, compressing the header according to the rule to produce a compressed packet of a predefined size that includes the compressed header and the payload; scheduling transmission of the compressed packet; and transmitting the compressed packet to a receiving device. In some embodiments, before compressing the frame, in response to determining that a size of the payload does not match a predefined size threshold: the payload is fragmented into a plurality of portions, wherein each portion satisfies the predefined size threshold, or the compressed packet is padded to the predefined size threshold via forward error correction padding information.

Abstract: Techniques are described herein for deploying, monitoring, and modifying network topologies comprising various computing and network nodes deployed across multiple workload resource domains. A deployment system may receive operational data from a network topology deployed across multiple workload resource domains, such as public or private cloud computing environments, on-premise data centers, and the like. The operational data may be provided to a trained machine-learning model, and output from the trained model may be used, along with constraint inputs and resource inventories of the workload resource domains, to determine updated topology models which may be deployed within the workload resource domains.

Abstract: Methods and systems for large silicon photonic interposers by stitching are disclosed and may include, in an optical communication system including a silicon photonic interposer, where the interposer includes a plurality of reticle sections: communicating an optical signal between first and second reticle sections utilizing a waveguide. The waveguide may include a taper region at a boundary between the two reticle sections, the taper region expanding an optical mode of the communicated optical signal prior to the boundary and narrowing the optical mode after the boundary. A continuous wave (CW) optical signal may be received in a first of the reticle sections from an optical source external to the interposer. The CW optical signal may be received in the interposer from an optical source assembly coupled to a grating coupler in the first of the reticle sections in the silicon photonic interposer.

Abstract: A network device receives multi-destination packets from a first node and forwards at least a first of the multi-destination packets to another network device using a first multi-destination tree with respect to the network device. The network device detects that a link associated with the first multi-destination tree satisfies one or more criteria and, in response to detecting that the link satisfies the one or more criteria, selects a second multi-destination tree with respect to the network device. The network device forwards at least a second of the multi-destination packets to the other network device using the second multi-destination tree.