Abstract: In one embodiment, a device clusters traffic feature vectors for a plurality of endpoints in a network into a set of clusters. Each traffic feature vector comprises traffic telemetry data captured for one of the endpoints. The device selects one of the clusters for labeling, based in part on contextual data associated with the clusters that was not used to form the clusters. The device obtains a device type label for the selected cluster by providing data regarding the selected cluster and the contextual data associated with that cluster to a user interface. The device provides the device type label and the traffic feature vectors associated with the selected cluster for training a machine learning-based device type classifier.

Abstract: Time Sensitive Networking (TSN) in wireless environments may be provided. First, a Radio Frequency (RF) profile associated with a station may be received by a computing device. Next, a number of Transmit Opportunities (TxOPs) to use for transmitting data between an Access Point (AP) and the station based on the received RF profile may be determined. The determined number of TxOPs may then be provided to a wireless controller associated with the AP.

Abstract: Techniques for improving the ability of FPGAs to process packets by implementing at least portions of the logic of packet parsers traditionally performed using the FPGA fabric as hardened resources, such as an Integrated Circuit (IC) block. The IC block receives bits of an incoming packet, carries these bits as a pipeline, and modifies a range of the bits through stages of aligners. The aligners extract header sections (or “windows”) of each packet header according to a shift amount, and the header sections are output to the FPGA fabric. The FPGA fabric includes extract and decision logic that maps the information included in the extracted header sections to a lookup vector, driving tables, and/or application logic. The FPGA provides shift amounts to subsequent aligners to cause the aligners to shift the packet bus such that previous header sections are removed.

Abstract: A method of streaming media content over a network from a media cache node is described. The method includes receiving a request for a media content item from a client device, the request comprising an address identifying a media content item to be streamed. In response to the request, a streaming engine process is allocated to the media content item for fulfilling the request. Based on the address identifying the media content item, a location comprising a media cache node able to provide the media content item is determined and the media content item is streamed to the client device using the streaming engine process allocated to the media content item. Further methods of streaming a media content item and providing access to media content are also described.

Abstract: In one embodiment, a service receives signal characteristic data indicative of characteristics of wireless signals received by one or more antennas located in a particular area. The service uses the received signal characteristic data as input to a Bayesian inference model to predict physical states of an object located in the particular area. A physical state of the object is indicative of at least one of: a mass, a velocity, an acceleration, a surface area, or a location of the object. The service updates the Bayesian inference model based in part on the predicted state of the object and a change in the received signal characteristic data and based in part by enforcing Newtonian motion dynamics on the predicted physical states.

Abstract: In one embodiment, a device in a network tracks changes in a source port or address identifier indicated by network traffic associated with a particular host in the network. The device detects an operating system start event based on the track changes in the source port or address identifier indicated in the traffic data associated with the particular host. The device provides data regarding the detected operating system start event as input to a machine learning-based malware detector. The device causes performance of a mitigation action in the network when the malware detector determines that the particular host is infected with malware.

Abstract: Various implementations disclosed herein include a thermal management system suitable electronic devices. In some implementations, a thermal management system includes an air guide and a cage wall, which together provide: an air intake having a first airflow area to produce a first air pressure; an airflow constriction, following the air intake, having a second airflow area smaller than the first airflow area to produce a second air pressure, and defining a first region; and an outlet following the airflow constriction having a third airflow area greater than the second airflow area to produce a third air pressure. The device cage has at least one aperture in the first region. The second air pressure is less than the first air pressure and the third air pressure and is sufficiently low to draw heated air from within the device cage through the at least one aperture to be expelled through the outlet.

Abstract: Optimized performance with a first media access protocol and a second media access protocol may be provided. First, media access for client devices associated with the first media access protocol may be scheduled for an Access Point (AP) for a first time period. The first time period may comprise a first predetermined amount of time. Next, media access for client devices associated with the first media access protocol may be paused for a second time period at the end of the first time period to allow client devices associated with the second media access protocol to access the media. The second time period may comprise a second predetermined amount of time. Then media access for client devices associated with the first media access protocol may be scheduled at the end of the second time period for a third time period. The third time period may comprise a third predetermined amount of time.

Abstract: Techniques for automated network adjustment are provided. A first data rate currently supported by a first access point is determined, and a second data rate is identified to be evaluated. A first sensor device of a plurality of sensor devices is identified, where a signal strength between the first sensor device and the first access point exceeds a predefined threshold. A first packet error rate is determined for communications between the first sensor device and the first access point using the first data rate, and a second packet error rate is determined for communications between the first sensor device and the first access point using the second data rate. Upon determining that the second packet error rate is lower than the first packet error rate, the first access point is automatically reconfigured to use the second data rate.

Abstract: Process margin relaxation is provided in relation to a compensated-for process via a first optical device, fabricated to satisfy an operational specification when a compensated-for process is within a first tolerance range; a second optical device, fabricated to satisfy the operational specification when the compensated-for process is within second tolerance range, different than the first tolerance range; a first optical switch connected to an input and configured to output an optical signal received from the input to one of the first optical device and the second optical device; and a second optical switch configured to combine outputs from the first optical device and the second optical device.

Abstract: This disclosure describes various methods, systems, and devices related to dynamic service node discovery in a network. In an example method, an intermediary node receives a Link Layer Discovery Protocol (LLDP) message from a first node. The LLDP message includes a discovery Type-Length-Value (TLV) that indicates a location of a service node in the network. The method further includes forwarding the LLDP message to a second node.

Abstract: Networked sleep mode management is provided by measuring network conditions for a first Access Point serving a plurality of client devices configured to operate in one of a sleep mode and an active mode; in response to detecting an amount of network usage devoted to transitioning members of the plurality of client devices from the sleep mode to the active mode satisfies a threshold: identifying a first subset of client devices from the plurality of client devices that are in the sleep mode; identifying a given client device from the first subset of client devices to transition to the active mode; and transmitting a tear-down message to the given client device that instructs the given client device to transition from the sleep mode to the active mode.

Abstract: Techniques for maintaining isolation and segregation for network paths through multi-cloud fabrics using VRF technologies. The techniques include running virtual routers in a cloud network that connect the cloud network to an on-premises network using a network overlay that preserves VRF information in data packets. Further, the virtual routers connect to individual gateways in the cloud network using tunnels, and each individual gateway is connected to multiple VPCs without overlapping subnets. The virtual routers may assign a sink VRF to each gateway connection that can be used to perform source-IP based VRF selection by mapping source IP addresses in each tunnel connection to appropriate VRFs for the source IP addresses. In this way, virtual routers may use sink VRFs to translate into the VRF information for data packets from the VPCs via source-IP based lookup, and use the corresponding VRF route table to determine next hops for data packets.

Abstract: This disclosure describes techniques for preventing message loops in communications among network devices. The techniques include preventing messages from being returned to a sending node and preventing messages from being sent more than once to any particular node. Such message loops may reverberate among network devices before triage efforts are able to stop the loop. As such, message loop prevention may help decrease the computational load among networked devices. At scale, the techniques may help prevent an exponential increase in data traffic that may propagate widely over the network. Loop prevention techniques may even prevent disabling of a network due to data traffic overload.

Abstract: In one embodiment, network operations are improved by performing updating operations data in an operations data field associated with the header of a particular protocol during the processing of a different protocol. A particular multiple-protocol (MP) packet is received by a particular network node in a network. The particular MP packet includes multiple protocol headers, including a first protocol header associated with a first protocol and a second protocol header associated with a second protocol. Further, the second protocol header associated with a second operations data field. During protocol processing of the first protocol on the particular MP packet, the second operations data field updated with particular operations data. The particular MP packet is sent from the particular network node, with said sent particular MP packet including said updated second operations data field with particular operations data.

Abstract: This disclosure describes techniques for scheduling workloads relating to data collection from network devices. The techniques may include weighting attributes that impact the data collection workload on computational, storage, and/or other resources of a data collector. The scheduling process may also include consideration of current usage rates at a data collector, to avoid overloading a particular collector while potentially underutilizing others. The weighting of attributes and/or consideration of usage rates may improve the efficiency of workload scheduling among a set of collectors. The overall result may include better utilization of the collectors and/or other resources of the computing network.

Abstract: A laser integrated photonic platform to allow for independent fabrication and development of laser systems in silicon photonics. The photonic platform includes a silicon substrate with an upper surface, one or more through silicon vias (TSVs) defined through the silicon substrate, and passive alignment features in the substrate. The photonic platform includes a silicon substrate wafer with through silicon vias (TSVs) defined through the silicon substrate, and passive alignment features in the substrate for mating the photonic platform to a photonics integrated circuit. The photonic platform also includes a III-V semiconductor material structure wafer, where the III-V wafer is bonded to the upper surface of the silicon substrate and includes at least one active layer forming a light source for the photonic platform.

Abstract: Techniques for utilizing Software-Defined Networking (SDN) controllers and network border leaf nodes of respective cloud computing networks to configure a data transmission route for a multicast group. Each border leaf node may maintain a respective external sources database, including a number of records indicating associations between a multicast data source, one or more respective border leaf nodes disposed in the same network as the multicast data source, and network capability information. A border leaf node, disposed in the same network as a multicast data source, may broadcast a local source discovery message to all border leaf nodes in remote networks to which it is communicatively coupled. A border leaf node may also communicate network capability information associated with one or more remote networks to a local SDN controller. The SDN controller may utilize the network capability information to configure a data transmission route to one or more destination nodes.

Abstract: In one embodiment, a network assurance system that monitors a network forms a cluster of similarly behaving wireless access points (APs). The cluster includes APs associated with different software versions. The network assurance system trains a machine learning-based failure prediction model for the cluster based on a set of features of the APs in the cluster. The network assurance system proactively triggers a client in the network to roam from a first AP to a second AP, based on the failure prediction model predicting a failure of the first AP. The network assurance system quarantines the failure prediction model when a new software version is associated with one or more of the APs.

Abstract: In one embodiment, a first server computing is configured to send, to a network component, a request to subscribe to a multicast group. In addition, the first server computing device may receive a message sent from a second server computing device to the multicast group. The first server computing device may determine, from the message, an internet protocol (IP) address and port number associated with the second server computing device. In addition, the first server computing device may determine that the first server computing device has not previously received information associated with the second server computing device. The first server computing device may cause a connection to be established between the first server computing device and the second server computing device, the connection for enabling the first server computing device to determine whether the second server computing device offers storage for the first server computing device.