Abstract: A method includes receiving in a network switch of a communication network communication traffic that originates from a source node and arrives over a route through the communication network traversing one or more preceding network switches, for forwarding to a destination node. In response to detecting in the network switch a compromised ability to forward the communication traffic to the destination node, a notification is sent to the preceding network switches. The notification is to be consumed by the preceding network switches and requests the preceding network switches to modify the route so as not to traverse the network switch.

Abstract: A method for communication includes transmitting data packets from a communication device to a network. Upon receiving in the communication device a congestion notification from the network, a rate of transmission of the data packets from the communication device to the network is reduced. While transmitting the data packets, after reducing the rate of transmission, the rate of transmission is increased incrementally when a predefined volume of data has been transmitted since having made a previous change in the rate of transmission.

Abstract: A method for communication in a packet data network that includes at least first and second subnets interconnected by multiple routers and having respective first and second subnet managers. The method includes assigning respective local identifiers to ports for addressing of data link traffic within each subnet, such that the first subnet manager assigns the local identifiers in the first subnet, and the second subnet manager assigns the local identifiers in the second subnet. The routers are configured by transmitting and receiving control traffic between the subnet managers and the routers. Data packets are transmitted between network nodes in the first and second subnets via one or more of the configured routers under control of the subnet managers.

Abstract: A communication network includes multiple nodes, which are arranged in groups such that the nodes in each group are interconnected in a bipartite topology and the groups are interconnected in a mesh topology. The nodes are configured to convey traffic between source hosts and respective destination hosts by routing packets among the nodes on paths that do not traverse any intermediate hosts other than the source and destination hosts.

Abstract: A method in a network element that forwards packets to destination nodes includes identifying groups of the destination nodes. Respective performance metrics of multiple different candidate network paths, over which the destination nodes in a given group are reachable from the network element, are estimated jointly for all the destination nodes in the given group. A network path is selected from among the candidate network paths based on the estimated performance metrics. The packets addressed to the destination nodes in the given group are forwarded over the selected network path.

Abstract: A method in a network element that includes multiple interfaces for connecting to a communication network includes receiving via an ingress interface packets that are not allowed to undergo re-routing and that are addressed to a destination via a first egress interface. The packets are forwarded via the first egress interface when there is a valid path from the first egress interface to the destination. When there is no valid path from the first egress interface to the destination, a second egress interface is selected from a group of multiple egress interfaces that have respective paths to the destination and are assigned to packets for which re-routing is allowed, and the packets are forwarded via the second egress interface until recovering a path to the destination.

Abstract: A method for communication includes configuring a multi-level fat-tree network to include at least three levels of switches, including multiple modules arranged externally in a tree topology. Each module contains a respective group of the switches arranged in an internal tree extending over at least two of the levels of the network. A subset of the modules is selected to be active in carrying the communication traffic. The network is operated so as to convey communication traffic among the switches via the active modules, while the modules that are not in the selected subset remain inactive.

Abstract: A method includes receiving in a network switch of a communication network communication traffic that originates from a source node and arrives over a route through the communication network traversing one or more preceding network switches, for forwarding to a destination node. In response to detecting in the network switch a compromised ability to forward the communication traffic to the destination node, a notification is sent to the preceding network switches. The notification is to be consumed by the preceding network switches and requests the preceding network switches to modify the route so as not to traverse the network switch.

Abstract: A method includes receiving in a network switch of a communication network communication traffic that originates from a source node and arrives over a route through the communication network traversing one or more preceding network switches, for forwarding to a destination node. In response to detecting in the network switch a compromised ability to forward the communication traffic to the destination node, a notification is sent to the preceding network switches. The notification is to be consumed by the preceding network switches and requests the preceding network switches to modify the route so as not to traverse the network switch.

Abstract: A method includes receiving in a network switch of a communication network communication traffic that originates from a source node and arrives over a route through the communication network traversing one or more preceding network switches, for forwarding to a destination node. In response to detecting in the network switch a compromised ability to forward the communication traffic to the destination node, a notification is sent to the preceding network switches. The notification is to be consumed by the preceding network switches and requests the preceding network switches to modify the route so as not to traverse the network switch.

Abstract: A method for communication in a packet data network that includes at least first and second subnets interconnected by multiple routers and having respective first and second subnet managers. The method includes assigning respective local identifiers to ports for addressing of data link traffic within each subnet, such that the first subnet manager assigns the local identifiers in the first subnet, and the second subnet manager assigns the local identifiers in the second subnet. The routers are configured by transmitting and receiving control traffic between the subnet managers and the routers. Data packets are transmitted between network nodes in the first and second subnets via one or more of the configured routers under control of the subnet managers.

Abstract: A method for communication includes sending communication packets over a network from a first network interface. A notification, which originates from a second network interface and indicates a network congestion encountered by one or more of the packets, is received in the first network interface. A network address of the second network interface is identified based on the notification. A transmission rate of subsequent packets addressed to the network address is regulated responsively to the notification, irrespective of a transport service instance on which the subsequent packets are sent from the first network interface.

Abstract: A method for communication includes configuring a multi-level fat-tree network to include at least three levels of switches, including multiple modules arranged externally in a tree topology. Each module contains a respective group of the switches arranged in an internal tree extending over at least two of the levels of the network. A subset of the modules is selected to be active in carrying the communication traffic. The network is operated so as to convey communication traffic among the switches via the active modules, while the modules that are not in the selected subset remain inactive.

Abstract: A method for communication includes estimating a characteristic of communication traffic to be carried by a fat-tree network. Responsively to the estimated characteristic, a subset of the spine switches in the highest level of the network is selected, according to a predetermined selection order, to be active in carrying the communication traffic. In each of the levels of the spine switches below the highest level, the spine switches to be active are selected based on the selected spine switches in a next-higher level. The network is operated so as to convey the traffic between the leaf switches via the active spine switches, while the spine switches that are not selected remain inactive.

Abstract: A method for communication includes sending communication packets over a network from a first network interface. A notification, which originates from a second network interface and indicates a network congestion encountered by one or more of the packets, is received in the first network interface. A network address of the second network interface is identified based on the notification. A transmission rate of subsequent packets addressed to the network address is regulated responsively to the notification, irrespective of a transport service instance on which the subsequent packets are sent from the first network interface.

Abstract: A method in a network element that forwards packets to destination nodes includes identifying groups of the destination nodes. Respective performance metrics of multiple different candidate network paths, over which the destination nodes in a given group are reachable from the network element, are estimated jointly for all the destination nodes in the given group. A network path is selected from among the candidate network paths based on the estimated performance metrics. The packets addressed to the destination nodes in the given group are forwarded over the selected network path.

Abstract: A method for communication includes estimating a characteristic of communication traffic to be carried by a fat-tree network. Responsively to the estimated characteristic, a subset of the spine switches in the highest level of the network is selected, according to a predetermined selection order, to be active in carrying the communication traffic. In each of the levels of the spine switches below the highest level, the spine switches to be active are selected based on the selected spine switches in a next-higher level. The network is operated so as to convey the traffic between the leaf switches via the active spine switches, while the spine switches that are not selected remain inactive.

Abstract: A device and a method for optimizing data transfer rate in an InfiniBand fabric is provided where a various number of transmitting devices aim data packets to a single receiving device or through a common link. The method which is implemented in an InfiniBand switch includes marking of packets in a rate corresponding to centrally configured marking rate, determination of the current number of data flows between the input ports and the output port of the switch and marking the data packet with Forward Explicit Congestion Notification according to an adaptive value of marking rate which depends on the initial value of the marking rate and is inversely proportional to the number of data flows.

Abstract: An embedded memory unit includes a memory array having a design size given by an integer n, the array including a matrix of memory cells arranged in n+1 lines, and selection circuitry, coupled to select n of the n+1 lines to which data are to be written and from which the data are to be read. A built-in self-test (BIST) circuit is coupled to test the memory array and to generate a binary pass/fail output. A repair machine is coupled to receive the binary output of the BIST circuit and to drive the selection circuitry to select different sets of n of the lines while driving the BIST circuit to test the memory array, so as to find one of the different sets for which the binary output indicates that the array has passed the self-test.

Abstract: An embedded memory unit includes a memory array having a design size given by an integer n, the array including a matrix of memory cells arranged in n+1 lines, and selection circuitry, coupled to select n of the n+1 lines to which data are to be written and from which the data are to be read. A built-in self-test (BIST) circuit is coupled to test the memory array and to generate a binary pass/fail output. A repair machine is coupled to receive the binary output of the BIST circuit and to drive the selection circuitry to select different sets of n of the lines while driving the BIST circuit to test the memory array, so as to find one of the different sets for which the binary output indicates that the array has passed the self-test.