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.

A distributed computing system, such as may be used to implement an electronic trading system, supports a notion of fairness in latency. The system does not favor any particular client. Thus, being connected to a particular access point into the system (such as via a gateway) does not give any particular device an unfair advantage or disadvantage over another. That end is accomplished by precisely controlling latency, that is, the time between when request messages arrive at the system and a time at which corresponding response messages are permitted to leave. The precisely controlled, deterministic latency can be fixed over time, or it can vary according to some predetermined pattern, or vary randomly within a pre-determined range of values.

A flow table management system can include a hardware memory module communicatively coupled to a network interface card. The hardware memory module is configured to store a flow table including a plurality of network flow entries. The network interface card further includes a flow table age cache configured to store a set of recently active network flows and a flow table management module configured to manage a duration for which respective network flow entries in the flow table stored in the hardware memory module remain in the flow table using the flow table age cache. In some implementations, age information about each respective flow in the flow table is stored in the hardware memory module in an age state table that is separate from the flow table.

A distributed computing system, such as may be used to implement an electronic trading system, supports a notion of fairness in latency. The system does not favor any particular client. Thus, being connected to a particular access point into the system (such as via a gateway) does not give any particular device an unfair advantage or disadvantage over another. That end is accomplished by precisely controlling latency, that is, the time between when request messages arrive at the system and a time at which corresponding response messages are permitted to leave. The precisely controlled, deterministic latency can be fixed over time, or it can vary according to some predetermined pattern, or vary randomly within a pre-determined range of values.

A network device processes received packets at least to determine port or ports of the network device via which to transmit the packet. The network device also classifies the packets into packet flows, the packet flows being further categorized into traffic pattern categories characteristic of traffic pattern characteristics of the packet flows. The network device buffers, according to the traffic pattern categories of the packet flows, packets that belong to the packet flows in a first packet memory or in a second packet memory, the first packet memory having a memory access bandwidth different from a memory access bandwidth of the second packet memory. After processing the packets, the network device retrieves the packets from the first packet memory or the second packet memory in which the packets are buffered, and forwards the packets to the determined one or more ports for transmission of the packets.

Embodiments of this application relate to the field of communications technologies, and disclose a dynamic scheduling method, an apparatus, and a system, so as to reduce information exchange costs and calculation complexity of data scheduling and route allocation. The method includes: receiving, by a scheduling platform, first VOQ length information of each aggregation switch in each timeslot; aggregating all the received first VOQ length information, to obtain global VOQ length information, where the global VOQ length information includes a total quantity of data packets that need to be sent from each of M pods to other pods than the pod; determining a transmission matching scheme based on the global VOQ length information; and sending corresponding matching result information to each aggregation switch according to the transmission matching scheme, so that each aggregation switch sends a data packet according to the transmission matching scheme.

The present systems and methods allow for rapid processing of large volumes of events. A producer node in a cluster determines a sharding key for a received event from an event stream. The producer node uses a sharding map to correlate the sharding key for the event with a producer channel, and provides the event to a producer event buffer associated with the producer channel. The producer event buffer transmits the event to a corresponding consumer event buffer associated with a consumer channel on a consumer node. The event processing leverages a paired relationship between producer channels on the producer node and consumer channels on the consumer node, so as to generate enhanced throughput. The event processing also supports dynamic rebalancing of the system in response to adding or removing producer or consumer nodes, or adding or removing producer or consumer channels to or from producer or consumer nodes.

Example methods and systems for adaptive polling. One example may comprise operating in a polling mode to poll, from a network interface, zero or more packets that require packet processing by the network device. The method may also comprise: in response to detecting a non-zero polling round, adjusting a polling parameter to delay switching from the polling mode to a sleep mode. The method may further comprise: in response to detecting a zero polling round and determining that a switch condition is satisfied, adjusting a sleep parameter associated with the sleep mode based on traffic characteristic information associated with one or more polling rounds; and switching from the polling mode to the sleep mode in which polling from the network interface is halted based on the sleep parameter.

A distributed computing system, such as may be used to implement an electronic trading system, supports a notion of fairness in latency. The system does not favor any particular client. Thus, being connected to a particular access point into the system (such as via a gateway) does not give any particular device an unfair advantage or disadvantage over another. That end is accomplished by precisely controlling latency, that is, the time between when request messages arrive at the system and a time at which corresponding response messages are permitted to leave. The precisely controlled, deterministic latency can be fixed over time, or it can vary according to some predetermined pattern, or vary randomly within a pre-determined range of values.

Techniques for transmitting data packets on a shared channel in a data communications network include storing, on a local node, a current number of turns in a transmitting queue and a current turn based on packets received from other nodes on the channel. For a first local data packet to transmit, a local transmit turn is obtained based on successfully transmitting the packet in a time interval following a last turn in the queue. A data link layer header includes queue fields for the current number of turns, the local transmit turn, and a request for adding the local transmit turn. While there is still a remaining packet to transmit and when the current turn is the local transmit turn, the packet is transmitted with queue fields indicating the current number of turns, the current turn, and an indication of no new turn. An acknowledgement control packet indicates successful transmission.