Described herein is a system and method for utilizing a protocol over RDMA network fabric between a first computing node and a second computing node. The protocol identifies a first threshold and a second threshold. A transfer request is received, and, a data size associated with the transfer request is determined. Based up the data size associated with the transfer request, one of at least three transfer modes is selected to perform the transfer request in accordance with the first threshold and the second threshold. Each transfer mode utilizes flow control and at least one RDMA operation. The selected transfer mode is utilized to perform the transfer request.

Embodiments relate to systems and methods that facilitate coordination amongst host devices that share network resources in order to use available bandwidth effectively. Embodiments can ensure the host devices themselves take responsibility for sending their data via shared network bandwidth, keeping access to bandwidth fair to all the host devices. Embodiments also include adapting to a continuously changing network bandwidth allocation policy for the shared network resource. In one embodiment, this includes using tokens (e.g., small chunk messages) to represent a grant to a host device to send a specified number of bytes over the network. Using a token generator module and a distributed queue, embodiments provide a unique and adaptive way to manage data transmissions among host devices within available resources.

The computer-implemented method for apportioning bandwidth in storage systems may include (i) identifying a plurality of storage media and at least one workload that is assigned a quantity of credits at the beginning of a predetermined time period that regulate a volume of bandwidth for input/output requests from the workload, (ii) detecting an input/output request from the workload, (iii) deducting, before fulfilling the input/output request, a number of credits from a current number of credits available to the workload based on an estimated quantity of bandwidth consumed by the input/output request, (iv) measuring an actual quantity of bandwidth consumed by the input/output request, and (v) adjusting the current number of credits available to the workload for the predetermined time period based on a difference between the estimated quantity of bandwidth and the actual quantity of bandwidth.

This application discloses a client service transmission method and apparatus. The method may include: receiving a client service, where the client service includes a plurality of data blocks, the client service is corresponding to a counter, and the counter is used to control an output rate of the client service; and sending the plurality of data blocks in a plurality of sending periods, where when a count value of the counter reaches a preset threshold in each sending period, at least one data block of the plurality of data blocks is sent. This technology may be applied to a scenario in which a transmission node transmits a client service.

A network switch capable of supporting cut-though switching and interface channelization with enhanced system performance. The network switch includes a plurality of ingress tiles, each tile including a virtual output queue (VOQ) scheduler operable to submit schedule requests to a fabric scheduler. Data is requested in unit of quantum which may aggregate multiple packets, which reduces schedule latency. Each request is associated with a start-of-quantum (SoR) state or a middle-of-quantum (MoR) state to support cut-through. The fabric scheduler performs a multi-stage scheduling process to progressively narrow the selection of requests, including stages of arbitration in virtual output port level, virtual output port group level, tile level, egress port level and port group level. Each tile receives the grants for its requests and accordingly sends request data to a switch fabric for transmission to the destination egress ports.

A technique allows stations to utilize an equal share of resources (e.g., airtime or throughput). This prevents slow stations from consuming too many resources (e.g., using up too much air time). Fairness is ensured by selective dropping after a multicast packet is converted to unicast. This prevents slow stations from using more than their share of buffer resources. Multicast conversion aware back-pressure into the network layer can be used to prevent unnecessary dropping of packets after multicast to unicast (1:n) conversion by considering duplicated transmit buffers. This technique helps achieve airtime/resource fairness among stations.

The present disclosure generally relates to controlling access to resources by selectively processing requests stored in a task queue to prioritize certain requests over others, thereby preventing automated scripts from accessing the resources. More specifically, the present disclosure relates to a normalization and prioritization system for controlling access to resources by queuing resource requests based on a client-defined normalization process that uses one or more data sources.

Methods and systems are provided for implementing a streaming deficit round robin arbiter to provide fair utilization of a single link. In some aspects, methods and systems are provided and can include specifying a quantum size indicating how much of a link of a stream is available for use, adding the quantum size to a deficit counter indicating available bandwidth, determining whether to provide a first data packet to an autonomous vehicle system based on the deficit counter and without determining a data packet size of the first data packet, and providing the first data packet to the autonomous vehicle system based on the determining of whether to provide the first data packet to the autonomous vehicle system.

Systems, methods, and computer-readable media are disclosed for an apparatus coupled to a communication bus, where the apparatus includes a queue and a controller to manage operations of the queue. The queue includes a first space to store a first information for a first traffic type, with a first flow class, and for a first virtual channel of communication between a first communicating entity and a second communicating entity. The queue further includes a second space to store a second information for a second traffic type, with a second flow class, and for a second virtual channel of communication between a third communicating entity and a fourth communicating entity. The first traffic type is different from the second traffic type, the first flow class is different from the second flow class, or the first virtual channel is different from the second virtual channel. Other embodiments may be described and/or claimed.

An integrated circuit includes queue circuits for storing packets, a scheduler circuit that schedules the packets received from the queue circuits to be provided in an output, and a traffic manager circuit that disables one of the queue circuits from transmitting any of the packets to the scheduler circuit based at least in part on a bandwidth in the output scheduled for a subset of the packets received from the one of the queue circuits.