Abstract:
Methods, apparatus and software for implementing enhanced data center congestion management for non-TCP traffic. Non-congested transmit latencies are determined for transmission of packets or Ethernet frames along paths between source and destination end-end-nodes when congestion along the paths is not present or minimal. Transmit latencies are similarly measured along the same source-destination paths during ongoing operations during which traffic congestion may vary. Based on whether a difference between the transmit latency for a packet or frame and the non-congested transmit latency for the path exceeds a threshold, the path is marked as congested or not congested. A rate at which the non-TCP packets are transmitted along the path is then managed as function of a rate at which the path is marked as congested. In one implementation, non-TCP traffic is managed by mimicking a Data Center TCP technique, under which the congestion marking status of the path is substituted as an input to a DCTP algorithm in place of the normally-used ECN-Echo flag input. The congestion window output by the DCTCP algorithm is then used to manage the rate at which non-TCP packets to be forwarded via the path are transmitted from a source end-node.
Abstract:
Methods and apparatus for implementing notification by network elements of packet drops. In response to determining a packet is to be dropped, a network element such as a switch or router determines the source of the packet and returns a dropped packet notification message to the source. Upon receipt of notification, networking software or embedded hardware on the source causes the dropped packet to be retransmitted. The notification may also be sent from the network element to the destination computer to inform networking software or embedded logic implemented by the destination computer that the packet was dropped and notification to the source has been sent, thus alleviating the destination from needing to send a Selective ACKnowledge (SACK) message to inform the source the packet was not delivered. (Too narrow)
Abstract:
Methods, apparatus and software for implementing enhanced data center congestion management for non-TCP traffic. Non-congested transit latencies are determined for transmission of packets or Ethernet frames along paths between source and destination end-end-nodes when congestion along the paths is not present or minimal. Transit latencies are similarly measured along the same source-destination paths during ongoing operations during which traffic congestion may vary. Based on whether a difference between the transit latency for a packet or frame and the non-congested transit latency for the path exceeds a threshold, the path is marked as congested or not congested. A rate at which the non-TCP packets are transmitted along the path is then managed as function of a rate at which the path is marked as congested. In one implementation, non-TCP traffic is managed by mimicking a Data Center TCP technique, under which the congestion marking status of the path is substituted as an input to a DCTP algorithm in place of the normally-used ECN-Echo flag input. The congestion window output by the DCTCP algorithm is then used to manage the rate at which non-TCP packets to be forwarded via the path are transmitted from a source end-node.
Abstract:
Method and apparatus for minimizing on-die memory in pull-mode switches. A disaggregated edge switch (DSW) includes a local communication interface including a plurality of outer ports configured to be coupled to a plurality of local hosts or servers via respective outer links, and a network interface configured to couple to at least one remote apparatus via a respective inner port and link. Data is transferred between the local hosts/servers using pull-mode transfers under which a receiver port “pulls” data from a local source port. The use of pull-mode data transfers between the local host/servers is managed by a scheduler so as to minimize the amount of buffers space for these local data transfers. Techniques are also disclosed for extending pull-mode data transfers using the Ethernet protocol over Ethernet links.
Abstract:
An embodiment may include circuitry that may provide, at least in part, at least one indication that at least one portion of data is available for processing by at least one data processor. The at least one indication may be provided, at least in part, prior to the entirety of the at least one portion of the data being available for the processing by the at least one data processor. The at least one data processor may begin the processing in response, at least in part, to the at least one indication. Many alternatives, variations, and modifications are possible.
Abstract:
Methods, apparatus and software for implementing enhanced data center congestion management for non-TCP traffic. Non-congested transmit latencies are determined for transmission of packets or Ethernet frames along paths between source and destination end-end-nodes when congestion along the paths is not present or minimal. Transmit latencies are similarly measured along the same source-destination paths during ongoing operations during which traffic congestion may vary. Based on whether a difference between the transmit latency for a packet or frame and the non-congested transmit latency for the path exceeds a threshold, the path is marked as congested or not congested. A rate at which the non-TCP packets are transmitted along the path is then managed as function of a rate at which the path is marked as congested. In one implementation, non-TCP traffic is managed by mimicking a Data Center TCP technique, under which the congestion marking status of the path is substituted as an input to a DCTP algorithm in place of the normally-used ECN-Echo flag input. The congestion window output by the DCTCP algorithm is then used to manage the rate at which non-TCP packets to be forwarded via the path are transmitted from a source end-node.
Abstract:
An embodiment may include circuitry that may provide, at least in part, at least one indication that at least one portion of data is available for processing by at least one data processor. The at least one indication may be provided, at least in part, prior to the entirety of the at least one portion of the data being available for the processing by the at least one data processor. The at least one data processor may begin the processing in response, at least in part, to the at least one indication. Many alternatives, variations, and modifications are possible.
Abstract:
Methods, apparatus, and networks configured to manage network congestion using packet recirculation. The networks employ network elements (e.g., Rbridges in Layer 2 networks and switches/routers in Layer 3 networks) that are configured to support multi-path forwarding under which packets addressed to the same destination may be routed via multiple paths to the destination. In response to network congestion conditions, such as lack of availability of a non-congested port via which a shortest path to the destination may be accessed, a packet may be routed backward toward a source node or forwarded toward a destination along a non-shortest path. The network elements may employ loopback buffers for looping packets back toward a source via the same link the packet is received on.