Mellanox Technologies, Ltd.

Mellanox Technologies Tlv Ltd.

Yuval Shpigelman

Roee Moyal

Yamin Friedman

Idan Burstein

Liron Mula

Omer Shabtai

Idan Borshteen

Gil Levy

Noam Bloch

Aviv Kfir

Lion Levi

Matty Kadosh

Niv Aibester

Tamir Ronen

Rotem Levinson

Liran Liss

Ariel Shahar

Eitan Zahavi

George Elias

Eyal Srebro

Haggai Eran

Michael Weiner

Sagi Kuks

Marian Pritsak

Alan Lo

Ofir Merdler

Yonatan Piasetsky

Yonatan Piasetzky

Kobi Pines

Charlie Mbariky

Amit Hermony

Shahrazad Hleihel

Reut Zuck

A network device includes one or more ports, processing circuitry, and a memory-network congestion controller. The one or more ports are to connect to a network. The processing circuitry is to run a plurality of processing tasks that access a shared memory, one or more of the processing tasks including communicating one or more packet flows over the network. The memory-network congestion controller is to identify a memory-access congestion, which occurs in accessing the shared memory by one or more of the processing tasks, and to alleviate the memory-access congestion by causing a reduction in a communication rate of at least one of the packet flows.

Alleviating memory-access congestion in network devices

Technologies for optimizing the spreading of traffic across multiple local output ports while considering both local load and end-to-end (E2E) load are described. One device has multiple outgoing ports and a network adapter that determines, for a first flow of packets, a first end-to-end (E2E) congestion rate of at least some of the outgoing ports. The network adapter determines a port state of at least some of the outgoing ports. The network adapter receives a first packet associated with the first flow of packets. The network adapter determines, using a first desired rate for the first flow, the first E2E congestion rates, and the port states, i) a first time at which the first packet is to be transmitted and ii) a first outgoing port on which the first packet is to be transmitted. The first packet is sent on the first outgoing port at the first time.

Combined congestion control and load balancing

A network device includes multiple ports, a Shared Buffer (SB) and a SB controller. The ports to connect to a communication network. The SB to temporarily store packets received from the communication network via the ports, the packets belonging to multiple flows. The SB controller to allocate one or more flow-specific storage regions in the SB, a given flow-specific storage region being allocated to store the packets that (i) belong to respective one or more of the flows and (ii) are to be transmitted via a respective egress queue. In response to detecting that an occupancy level in the given flow-specific storage region exceeds a specified occupancy threshold, the SB controller to report the flows in the given flow-specific storage region as congested.

Flow-based congestion control

Technologies for spreading packets of transport flows across multiple network paths are described. A network controller includes a transport layer and a network layer. The transport layer includes a flow scheduler to schedule a transport flow from one of a plurality of transport flows. The network layer includes multipath logic to receive packets from the transport flow and select which path of a plurality of paths to a destination to use for the packets based on path congestion weights corresponding to the plurality of paths.

Routing transport flows in a transport layer over multiple paths in a network layer

Technologies for spreading a burst of data across multiple network paths in remote direct memory access (RDMA) over converged Ethernet (RoCE) and InfiniBand are described. A network interface controller sends a first burst of a transport flow directed to a second node over a first network path. The network interface controller determines that a second burst is to be sent over a different network path, and identifies a second network path using a multipath context. The multipath context stores a first weight value or a first state associated with the first network path and a second weight value or a second state associated with the second network path. The network interface controller sends the second burst of data to the second node via the second network path.

Remote direct memory access (RDMA) multipath

In one embodiment, data communication device includes a network interface to receive first packets over a network from another network device via a switch, which includes a buffer associated with a variable buffer delay, and packet processing circuitry to compute respective measures of delay over the network to the other network device over time responsively to the received first packets, find a minimum measure of delay over the network to the other network device responsively to at least some of the computed respective measures of delay, estimate a current measure of buffer delay of the buffer responsively to the found minimum measure of delay and a current one of the computed respective measures of delay, set a packet processing parameter responsively to the estimated current measure of buffer delay, and process second packets responsively to the set packet processing parameter.

Network delay estimation

A network element includes a transmit-queue for transmitting packets from at least two sources, each source having a predefined priority level, to a headroom buffer in a peer network element. Flow-control circuitry receives from the peer network element signaling that indicates a number of credits for transmitting packets to the peer network element, manages a current number of credits available for transmission from the transmit-queue, responsive to the signaling, selects a threshold priority based on the current number of credits for the transmit-queue; and transmits packets associated with data sources of the transmit-queue that are higher in priority than the threshold priority, and refrain from transmitting other packets associated with the transmit-queue.

Head-of-queue blocking for multiple lossless queues

A network adapter includes a port and one or more circuits. The port communicates packets over a network in which switches forward packets in accordance with tuples of the packets. The one or more circuits are to hold a user-programmable scheme specifying assignments of the packets of a given flow destined to a peer node to sub-flows having respective different tuples, assign first packets of the given flow to one or more of the sub-flows in accordance with the user-programmable scheme, by setting respective tuples of the first packets, transmit the first packets to the peer node via the port, monitor notifications received from the network, the notifications being indicative of respective states of the sub-flows, based on the notifications and on the user-programmable scheme determine an assignment of second packets of the given flow to the sub-flows, and transmit the second packets to the peer node via the port.

Flexible per-flow multipath managed by sender-side network adapter

A device and method may alter the transmission rate of data sent across a computer network based on a time to receive an acknowledgement in response to a packet sent over the network. An embodiment may transmit packets across the computer network according to a rate R, where R is determined based at least on a number of bytes to be sent during a window (cwnd) divided by a duration of time (RTT); and modify RTT based on a current round trip time of a packet sent over the network (e.g. based on a time to receive an acknowledgement in response to a packet sent over the network).

System and method for network rate limiting

Technologies for spreading a single transport flow across multiple network paths in remote direct memory access (RDMA) over converged Ethernet (RoCE) and InfiniBand are described. A network interface controller receives a first packet and a second packet of a transport flow directed to a second node. The network interface controller assigns a first network routing identifier to the first packet and a second network routing identifier to the second packet, the first network routing identifier corresponding to a first network path between the first and second nodes, the second network routing identifier corresponding to a second network path between the first node and the second node. The network interface controller schedules a first packet of the transport flow to be sent via the first network path and a second packet of the transport flow to be sent via the second network path.

Devices, methods, and systems are provided. In one example, a method is described to include measuring a cardinality of actual data flows at a flow-processing resource, determining that the cardinality of the actual data flows triggers a congestion control action, and, in response to determining that the cardinality of the actual data flows triggers the congestion control action, implementing the congestion control action with respect to the flow-processing resource.

Cardinality-based traffic control

An apparatus includes a network interface and a processor. The network interface communicates with a network including switches interconnected in a Cartesian topology having multiple dimensions. The processor predefines turn types of turns in the Cartesian topology, each turn traverses first and second hops along first and second dimensions having same or different respective identities, and each turn type is defined at least by identities of the first and second dimensions. The processor searches for a preferred route from a source switch to a destination switch, by evaluating candidate routes based on the number of VLs required for preventing a deadlock condition caused by the candidate route. The number of VLs required depends on a sequential pattern of turn types formed by the candidate route. The processor configures one or more switches in the network to route packets from the source switch to the destination switch along the preferred route.

Producing deadlock-free routes in lossless cartesian topologies with minimal number of virtual lanes

In one embodiment, a system includes a peripheral device including a memory access interface to receive from a host device headers of packets, while corresponding payloads of the packets are stored in a host memory of the host device, and descriptors being indicative of respective locations in the host memory at which the corresponding payloads are stored, a data processing unit memory to store the received headers and the descriptors without the payloads of the packets, and a data processing unit to process the received headers, wherein the peripheral device is configured, upon completion of the processing of the received headers by the data processing unit, to fetch the payloads of the packets over the memory access interface from the respective locations in the host memory responsively to respective ones of the descriptors, and packet processing circuitry to receive the headers and payloads of the packets, and process the packets.

Zero-copy processing

A network element includes circuitry and multiple ports. The multiple ports are configured to connect to a communication network. The circuitry is configured to receive via one of the ports a packet that originated from a source node and is destined to a destination node, the packet including a mark that is indicative of a cumulative state derived from at least bandwidth utilization conditions of output ports that were traversed by the packet along a path, from the source node up to the network element, to select a port for forwarding the packet toward the destination node, to update the mark of the packet based at least on a value of the mark in the received packet and on a local bandwidth utilization condition of the selected port, and to transmit the packet having the updated mark to the destination node via the selected port.

Explicit notification of operative conditions along a network path

A network device includes processing circuitry and a plurality of ports. The ports connect to a communication network. The processing circuitry is configured to receive, via an input port, data packets and probe packets that are addressed to a common output port, to store the data packets in a first queue and the probe packets in a second queue, both the first queue and the second queue are served by the output port, to produce telemetry data indicative of a state of the network device, based on a processing path that the data packets traverse within the network device, to schedule transmission of the data packets from the first queue at a first priority, and schedule transmission of the probe packets from the second queue at a second priority higher than the first priority, and to modify the scheduled probe packets so as to carry the telemetry data.

Growing community of inventors

Netanya, Israel

Yuval Shpigelman