Mellanox Technologies Tlv Ltd.

Mellanox Technologies, Ltd.

Alex Shpiner

Gil Levy

Matty Kadosh

Liran Liss

Eitan Zahavi

Benny Koren

Noam Bloch

Barak Gafni

Liron Mula

Idan Burstein

Zachy Haramaty

Dror Goldenberg

Roee Moyal

Shachar Raindel

Adi Menachem

Vladimir Zdornov

Sagi Kuks

Dror Bohrer

Tal Anker

A network adapter includes a host interface and circuitry. The host interface is configured to connect locally between the network adapter and a host via a bus. The circuitry is configured to receive from one or more source nodes, over a communication network to which the network adapter is coupled, multiple packets destined to the host, and temporarily store the received packets in a queue of the network adapter, to send the stored packets from the queue to the host over the bus, to monitor a performance attribute of the bus, and in response to detecting, based at least on the monitored performance attribute, an imminent overfilling state of the queue, send a congestion notification to at least one of the source nodes from which the received packets originated.

Managing congestion in a network adapter based on host bus performance

A method for communication includes, in a first network switch that is part of a communication network having a topology, detecting a compromised ability to forward a flow of packets originating from a source endpoint to a destination endpoint. In response to detecting the compromised ability, the first network switch identifies, based on the topology, a second network switch that lies on a current route of the flow, and also lies on one or more alternative routes from the source endpoint to the destination endpoint that do not traverse the first network switch. A notification, which is addressed individually to the second network switch and requests the second network switch to reroute the flow, is sent from the first network switch.

Unicast forwarding of adaptive-routing notifications

A network element includes multiple ports and packet processing circuitry. The ports are configured for exchanging packets with a communication network. The packet processing circuitry is configured to forward first packets over a forward path from a source node to a destination node, to forward second packets over a reverse path, which is opposite in direction to the forward path, from the destination node to the source node, and to mark one or more of the second packets that are forwarded over the reverse path, with an indication that notifies the source node that congestion is present on the forward path.

Switch-enhanced short loop congestion notification for TCP

A network device includes circuitry and multiple ports. The circuitry is configured to hold a definition of a normalization function that determines, based on (i) a reference probability of applying a congestion indication operation to packets having a predefined reference packet-size and (ii) a packet-size parameter, a normalized probability of applying the congestion indication operation to packets whose size equals the packet-size parameter. The normalization function depends exponentially on a ratio between the packet-size parameter and the reference packet-size. The circuitry id configured to store packets in a queue, and to schedule transmission of at least some of the queued packets via an output port, to calculate the normalized probability for a given packet, by applying the normalization function to an actual reference probability and an actual size of the given packet, and randomly apply a congestion indication operation to the given packet, in accordance with the normalized probability.

Probabilistic normalized congestion indication based on packet size

A network switch includes switch circuitry and multiple ports. The ports are configured to communicate with a communication network. The switch circuitry is configured to receive via the ports multiple packets, which are destined to a destination network node and which specify attributes used by the destination network node as cache keys for on-demand fetching of context items into a cache memory of the destination network node, to control a rate of fetching the context items into the cache memory at the destination network node, by ordering the received packets in a sequence, based on the attributes of the respective packets, using an ordering criterion that aims to place packets that access a common context item in proximity to one another in the sequence, and to forward the received packets to the destination network node, via the ports, in accordance with the ordered sequence.

Packet scheduling in a switch for reducing cache-miss rate at a destination network node

A network element connected to a data network holds a flow of data packets in a queue and periodically determines a metric of the queue. Responsively to a predetermined value of the metric the queue is associated with an elephant flow or a mouse flow. The packets are marked according to the associated flow, and the network element sends the marked packets into the data network. Other network elements process the packets according to the associated flow marked therein.

Elephant flow detection in network access

A processor is configured to receive, from a client, a first message indicating a request to establish a connection between the client and a server, to ascertain that the first message does not include any cookie satisfying one or more criteria, to send, to the client, a second message that includes a first cookie, without allocating an endpoint on the server for the connection, in response to ascertaining that the first message does not include any cookie satisfying the criteria, to receive subsequently, from the client, a third message, to ascertain that the third message includes a second cookie, and that the second cookie satisfies the criteria, to allocate the endpoint for the connection in response to ascertaining that the second cookie satisfies the criteria, and to send, to the client, a fourth message indicating that the server is ready to receive data communication at the allocated endpoint.

Defending against DoS attacks over RDMA connections

A method for network communication includes receiving in a network element a packet for forwarding to a destination node. The destination node is reachable via two or more candidate ports of the network element that are connected to respective next-hop network elements. Link-level flow-control credit notifications are received in the network element from the next-hop network elements via the respective candidate ports. An egress port is selected for the packet, from among the candidate ports, based at least on the received link-level flow-control credit notifications. The packet is forwarded toward the destination node over the selected egress port.

Adaptive routing based on flow-control credits

Communication apparatus includes multiple interfaces configured to be connected to respective links in a packet data network. Switching circuitry in the apparatus is coupled between the interfaces and is configured to receive, via a first interface among the multiple interfaces, an adaptive routing notification (ARN) requesting that a specified flow of packets from a given source to a given destination in the network be rerouted. The switching circuitry is configured, upon verifying that the first interface serves as an egress interface for the packets in the specified flow, to reroute the specified flow through a different, second interface among the multiple interfaces when there is an alternative route available in the network from the second interface to the given destination, and after finding that there is no alternative route available from any of the interfaces to the given destination, to forward the ARN to a plurality of the interfaces.

Forwarding of adaptive routing notifications

Communication apparatus includes multiple interfaces for connection to a packet data network and a memory configured to contain packets awaiting transmission to the network in multiple queues, which are assigned respective transmission priorities. Control logic assigns to the queues respective weighting factors, which vary inversely with the respective transmission priorities, and calculates for each egress interface a respective interface congestion level. The control logic calculates effective congestion levels for the queues as a weighted function of the respective queue lengths and the respective interface congestion level, weighted by the respective weighting factors, and applies congestion control to the queues responsively to the effective congestion levels.

Congestion estimation for multi-priority traffic

Communication apparatus includes a plurality of interfaces configured to be connected to a Layer-3 packet network and to serve as ingress and egress interfaces to receive and transmit packets from and to the network. Routing logic is coupled to process respective Layer-3 headers of the packets received through the ingress interfaces and to route the packets via the egress interfaces to respective destinations indicated by the Layer-3 headers. Congestion detection logic is coupled to identify a flow of the received packets that is causing congestion in the network and a Layer-3 address from which the flow originates, and to direct the routing logic to route a backward congestion notification message (CNM) packet via one of the egress interfaces to the identified Layer-3 address.

Growing community of inventors

Nesher, Israel

Alex Shpiner