Cray UK Limited

Gnodal Limited

Cray Hk Limited

Jon Beecroft

David Charles Hewson

Anthony Michael Ford

Edward James Turner

Mark Owen Homewood

A multi-path network for use in a bridge, switch, router, hub or the like, includes a plurality of network ports; a plurality of network elements; and a plurality of network links interconnecting the network elements and the network ports, for transporting data packets. Each network element has a fault monitor for detecting faults on the network links to which the network element is connected, a plurality of dynamically selectable output ports and output port selection mechanism. Each network element also being adapted to communicate the existence of a fault back to one or more other network elements so that network elements connected to the faulty network link can be bypassed, and each network element being adapted to update the output port selection mechanism when communication of the existence of a fault is received so that only those output ports which ensure the faulty network link is bypassed are dynamically selectable. Also, a method of managing faults in a multi-path network utilizes a similar methodology to insure that identified faults are avoided.

Multi-path network with fault detection and dynamic adjustments

A method of controlling data packet congestion in a data packet network comprises determining a reference flow condition that results in data packet congestion at a node of a data packet network, and identifying a data packet flow having a flow condition substantially equal to the reference flow condition. For such an identified data packet flow the following actions are taken causing a data packet to be dropped from the identified data packet flow, allowing a predetermined number of data packets from the identified data packet flow to proceed, and dropping data packets from the identified data packet flow subsequent to the predetermined number of data packets, until the packets that were not dropped have been delivered to the egress ports of the network.

Apparatus and method

The present invention relates to data switching networks, and, in particular, to link aggregation groups in Ethernet switching networks. A technique is described in which a large number of links in a link aggregation group can be managed.

Ethernet switch and method for routing Ethernet data packets

The present invention provides a method of preserving packet ordering in a multipath network having a plurality of network elements interconnected by network links wherein for each data packet arriving at an egress port of the multi-path network, a delivery acknowledgement is issued by the egress port and is transmitted across the network following in reverse the path taken by the data packet being acknowledged. The state of each link in the path taken by the data packet being acknowledged is updated by the acknowledgement. The present invention further provides a multi-path network for use in a bridge, switch, router, hub or the like, the multi-path network comprising a plurality of network ports; a plurality of network elements; and a plurality of network links interconnecting the network elements and the network ports for transporting data packets, each network egress port including an acknowledgement mechanism for issuing a delivery acknowledgement in response to receipt of a data packet and each network element being adapted to transmitted a delivery acknowledgment in the opposite direction along the path taken by the data packet being acknowledged and being further adapted to update the state of at least one of the network links to which it is connected in response to receipt of an acknowledgement. The invention further provides an Ethernet bridge or router incorporating such a multi-path network.

Method of data delivery across a network

The present invention relates to a method of managing congestion in a multi-path network, the network having a plurality of network elements arranged in a plurality of switch stages and a plurality of network links interconnecting the network elements, the method comprising the steps of detecting congestion on a network link, the congested network link interconnecting the output port of a first network element with a first input port of a second network element in a subsequent switch stage; identifying an uncongested network link connected to a second input port of said second network element; and directing future data packets on a route across the multi-path network which includes the identified uncongested network link. Also provided is a multi-path network and an Ethernet bridge or router incorporating such a multi-path network.

A data processing system comprises a first clock domain having a first clock rate, a second clock domain having a second clock rate, and a data path operable to transfer data items from the first clock domain to the second clock domain. The data path comprises a buffer having an input for receiving data items from the first clock domain, and an output port for transmitting data items to the second clock domain in a first-in first-out manner. The buffer has a first pointer for indication of a current first location of the buffer, and a second pointer for indication of a current second location of the buffer. The system further includes a read controller operable to define a read pattern for the buffer, to control output from the buffer in dependence upon such a read pattern, and to adjust such a read pattern in dependence upon a value of such a first pointer for the buffer.

Synchronizing data transfer between a first clock domain and second clock domain

The invention provides an Ethernet bridge or router comprising a network fabric adapted to provide interconnectivity to a plurality of Ethernet ports, each of the Ethernet ports being adapted to receive and/or transmit Ethernet frames, and wherein the Ethernet bridge or router further comprises an encapsulator connected to receive Ethernet Protocol Data Units from the Ethernet ports, wherein the encapsulator is operable to generate a Fabric Protocol Data Unit from a received Ethernet Protocol Data Unit, the Fabric Protocol Data Unit comprising a header portion, and a payload portion which comprises the Ethernet Protocol Data Unit concerned, and wherein the encapsulator is operable to transform Ethernet destination address information from the Ethernet Protocol Data Unit into a routing definition for the network fabric, and to include this routing definition in the header portion of the Fabric Protocol Data Unit. Also provided is a method of data delivery across a network.

Method of data delivery across a network fabric in a router or Ethernet bridge

A network switch adaptive routing mechanism is disclosed along with an associated method of data routing. In particular, a mechanism is disclosed that enables adaptive routing selection within a crossbar switch.

Adaptive routing apparatus and method

A method and system is provided for efficiently controlling data packet routing through a switching element which has a first plurality of input ports and a second plurality of output ports. The method and system are configured for detecting issuance of a first congestion indicator from an output port of the switching element which is related to a first routing request received from a first input port. Reassertion of the first routing request is then carried out after a first time period. Similarly, the method and system are configured for detecting issuance of a second congestion indicator from the output port of the switching element, which is related to a second routing request from a second input port. Reassertion of the second routing request is then carried out after a second time period. In addition, the first and second time periods are set up so that reassertions of the first and second connection requests occur substantially simultaneously.

Method and device

The present invention relates to a method of sorting data packets in a multi-path network having a plurality of ports; a plurality of network links; and a plurality of network elements, each network element having at least first and second separately addressable buffers in communication with a network link and the network links interconnecting the network elements and connecting the network elements to the ports, the method comprising: sorting data packets with respect to their egress port or ports such that at a network element a first set of data packets intended for the same egress port are queued in said first buffer and at least one other data packet intended for an egress port other than the egress port of the first set of data packets is queued separately in said second buffer whereby said at least one other data packet is separated from any congestion associated with the first set of data packets. The present invention further relates to a method of data delivery in a multi-path network comprising the sorting of data packets in accordance with a first aspect of the present invention. The present invention further relates to a multi-path network operable to sort data packets and operable to deliver data in a multi-path network.

The present invention provides a multi-path network for use in a bridge, switch, router, hub or the like, comprising a plurality of network ports adapted for connection with one or more devices, each device having a different identifying address data; a plurality of network elements; and a plurality of network links interconnecting the network elements and connecting the network elements to the network ports, wherein the multi-path network further comprises separately addressable memory elements each adapted for storing device address data and the multi-path network is adapted to distribute a plurality of device address data amongst the plurality of memory elements.

Multi-path network

The present invention provides a method of preserving packet ordering in a multi-path network having a plurality of network elements interconnected by network links wherein for each data packet arriving at an egress port of the multi-path network, a delivery acknowledgement is issued by the egress port and is transmitted across the network following in reverse the path taken by the data packet being acknowledged, and wherein the state of each link in the path taken by the data packet being acknowledged is updated by the acknowledgement. The present invention further provides a multi-path network for use in a bridge, switch, router, hub or the like, the multi-path network comprising a plurality of network ports; a plurality of network elements; and a plurality of network links interconnecting the network elements and the network ports for transporting data packets, each network egress port including acknowledgement means for issuing a delivery acknowledgement in response to receipt of a data packet and each network element being adapted to transmitted a delivery acknowledgment in the opposite direction along the path taken by the data packet being acknowledged and being further adapted to update the state of at least one of the network links to which it is connected in response to receipt of an acknowledgement. The invention further provides an Ethernet bridge or router incorporating such a multi-path network.

The present invention provides a method of limiting the frequency of floods within a data network, the floods arising as a data frame is routed to an unknown destination, the method comprising the steps of: receiving on an ingress port a data frame intended for a destination station and containing a MAC address of that station; checking the destination MAC address with the contents of a MAC table; and thereby determining whether the data frame is to be routed, discarded or flooded to all ports except the ingress port; and respectively routing, discarding or flooding the data frame to all ports except the ingress port. The present invention further provides a method of populating a MAC table within an Ethernet network with entries relating a MAC station address to a port to which it is connected, the method comprising the steps of: receiving on an ingress port a data frame intended for a destination station and containing a MAC address of that station; checking the destination MAC address with the contents of a MAC table; and thereby determining whether the destination MAC address and ingress port are to be written to the MAC table and, if so writing the destination MAC address and ingress port to the MAC table; and flooding the data frame to all ports except the ingress port.

Growing community of inventors

Bristol, United Kingdom

Jon Beecroft