A graphics processing unit (GPU) includes a packet management component that automatically aggregates data from input packets. In response to determining that a received first input packet does not indicate a send condition, and in response to determining that a generated output packet would be smaller than an output size threshold, the packet management component aggregates data corresponding to the first input packet with data corresponding to a second input packet stored at a packet buffer. In response to determining that a received third input packet indicates a send condition, the packet management component sends the aggregated data to a compute unit in an output packet and performs an operation indicated by the send condition.

If a communication apparatus is to transmit data to another communication apparatus and communication via a communication unit included in the other communication apparatus is not performable, whether or not to transmit a frame for causing a transition to a state where the communication via the communication unit included in the other communication apparatus is performable is selected based on an amount of data accumulated in a transmission queue in which the data is stored.

This application discloses a packet forwarding method and apparatus, a system, a device, and a storage medium. The method is applied to an in-vehicle network and includes: obtaining, by a network device, a plurality of packets carried on an Ethernet link in the in-vehicle network, where the plurality of packets are packets of one or more application categories; determining a traffic class of each packet based on an application category of each packet in the plurality of packets; adding each packet to a forwarding queue corresponding to the traffic class based on the traffic class of each packet; and scheduling and forwarding packets in all forwarding queues based on an absolute priority and a traffic class corresponding to each forwarding queue, and using a shaper in a forwarding queue of at least one traffic class in intermediate traffic classes in the scheduling and forwarding process.

An access control unit includes packet buffers provided for each of users, a packet identification unit that stores received packets in a corresponding packet buffer, a scheduling unit that decides a packet buffer to be the object of transfer, a transfer control unit that, in a case that updating of reference data can be performed at an application processing circuit, and also the packet buffer decided by the scheduling unit is different from the current packet buffer that is the object of transfer, updates to reference data corresponding to the packet buffer decided by the scheduling unit, and a buffer selection unit that connects the packet buffers decided to be the object of transfer to the packet transfer unit when updating of reference data is completed.

A communication system in which DRA control is aided by RL. An example embodiment may control one or more buffer queues populated by downstream and/or upstream data streams. The egress rates of the buffer queues can be dynamically controlled using an RL technique, according to which a learning agent can adaptively change the state-to-action mapping function of the DRA controller while circumventing the RL exploration phase and relying on extrapolation of the already taken actions instead. This feature may result in at least two benefits: (i) cancellation of a performance penalty typically associated with RL exploration; and (ii) faster learning of the environment, as the learning agent can determine the performance metrics of many actions per state in a single occurrence of the state. In an example embodiment, the communication system may be a DSL system, a PON system, or a wireless communication system.

A network device, including ports that receive/send data packets from/to a network, receives data packets of multiple traffic flows, and populates a queue in memory with the data packets. The network device periodically updates a fair rate for the multiple traffic flows to converge a length of the queue to a reference length. Specifically, the network device determines a length of the queue, a change in the length from a previous length, and a deviation of the length from the reference length. The network device detects an increase in the change in length above a threshold that is based on the reference length. If the increase is not above the threshold, the network device derives the fair rate from a previous fair rate using proportional integral control. The network device identifies elephant flows among the multiple traffic flows, and sends the fair rate to a source of each elephant flow.

A method for managing several queues of a network interface card (NIC) of a computer. The method initially configures the NIC to direct data messages received for a data compute node (DCN) executing on the computer to a default first NIC queue. When the DCN requests data messages addressed to the particular DCN to be processed with a first feature for load balancing data messages across multiple queues and a second feature for aggregating multiple related data messages into a single data message, the method configures the NIC to direct subsequent data messages received for the DCN to a second queue in a first subset of queues associated with the first feature if a load on the default first queue exceeds a first threshold. Otherwise, if a load on the first subset of queues exceeds a second threshold, the method configures the NIC to direct subsequent data messages received for the particular DCN to a third queue in a second subset of queues associated with both the first and second features.

Systems and methods of using a packet order work scheduler (POWS) to assign packets to a set of scheduler queues for supplying packets to parallel processing units. A processing unit and the associated scheduler queue are dedicated to a specific flow until a queue-reallocation event, which may correspond to the associated scheduler queue being idle for at least a certain interval as indicated by its age counter, or the queue being the least recently used, when a new flow arrives. In this case, the scheduler queue and the associated processing unit may be reallocated to the new flow and disassociated with the previous flow. As a result, dynamic packet workload balancing can be advantageously achieved across the multiple processing paths.

Systems and methods are provided for managing a data communication within a multi-level network having a plurality of switches organized as groups, with each group coupled to all other groups via global links, including: at each switch within the network, maintaining a global fault table identifying the links which lead only to faulty global paths, and when the data communication is received at a port of a switch, determine a destination for the data communication and, route the communication across the network using the global fault table to avoid selecting a port within the switch that would result in the communication arriving at a point in the network where its only path forward is across a global link that is faulty; wherein the global fault table is used for both a global minimal routing methodology and a global non-minimal routing methodology.

Embodiments relate to the management of data traffic congestion in a network communication node, the network communication node comprising a queue buffer configured to respectively enqueue packets at an input and dequeue packets at an output, and an Active Queue Management (AQM) module configured to determine a drop or a mark decision for a packet based on control parameters, wherein values for the control parameters are derived based on values of queue parameters weighted with respective weight factors and their associated target values, values of queue parameters and their associated target values weighted with respective weight factors, or a combination thereof.