Cisco Technology, Inc.

Steven Howard Berl

Ulrica Tam

Saravanan Agasaveeran

Qingming Ma

Zhanhe Shi

John Gary Waclawsky

Leo Pereira

Dirk Anteunis

Kushal A Patel

Arnold Stamler

Robert Charles Monsen

Dario Calia

Minjie Lin

David Scott McCowan

Satish Gannu

Chris Riggins

James Balestriere

Aniruddha Khare

Gordon Cox

M Salim Olime

Rajesh Agrawal

John David Billings

Controlling congestion in a networking device having a plurality of input interface queues comprises estimating, in each of one or more sampling states, a data arrival rate for each of the plurality of input interface queues with respect to incoming data packets received on corresponding input interfaces, obtaining a set of estimated arrival rates for the plurality of the input interface queues, determining, for each polling state associated with a respective sampling state, the sequence in which the plurality of input interface queues should be polled using the set of estimated data arrival rates of the plurality of input interface queues, and polling the plurality of interface queues in accordance with the determined sequence. The sequence indicates when, during a single polling cycle, each of the input interface queues should be polled in relation to every other of the input interface queues.

Apparatus and method for rate-based polling of input interface queues in networking devices

A method of automatically configuring a port of a network element comprises the steps of discovering information that identifies or describes a second network element that is coupled to a port of a first network element; associating the port with a port role definition selected from a plurality of port role definitions based on the discovered information; retrieving one or more configuration settings that are associated with the selected port role definition; and applying the one or more configuration settings to the port.

Method and apparatus providing role-based configuration of a port of a network element

The congestion control in a networking device having a plurality of input interface queues includes (a) estimating the data arrival rate on each of the plurality of input interface queues, and (b) determining, for each polling round, the sequence in which the plurality of input interface queues should be polled and the quantity of data to be processed from each of the plurality of input interface queues each time the input interface queue is polled, using the estimated data arrival rate on each of the plurality of input interface queues.

A two-phase packet processing technique is provided for routing traffic in a packet-switched, integrated services network which supports a plurality of different service classes. During Phase I, packets are retrieved from the router input interface and classified in order to identify the associated priority level of each packet and/or to determine whether a particular packet is delay-sensitive. If it is determined that a particular packet is delay-sensitive, the packet is immediately and fully processed. If, however, it is determined that the packet is not delay-sensitive, full processing of the packet is deferred and the packet is stored in an intermediate data structure. During Phase II, packets stored within the intermediate data structure are retrieved and fully processes. The technique of the present invention significantly reduces packet processing latency, particularly with respect to high priority or delay-sensitive packets. It is easily implemented in conventional routing systems, imposes little computational overhead, and consumes only a limited amount of memory resources within such systems.

Packet prioritization processing technique for routing traffic in a packet-switched computer network

A mechanism conveys information pertaining to transmission priority (TP) levels of inbound packets transmitted over a heterogeneous network from a switching node to a hybrid node of the network. The mechanism comprises a packet-recognizing filter having a format that is generated by the hybrid node and dynamically transmitted to the switching node over a predefined communication channel of the network. The filter enables the switching node to classify the inbound packets and assign them appropriate TP levels.

Mechanism for conveying data prioritization information among heterogeneous nodes of a computer network

A memory usage data structure (MUDS) is maintained for each process executing in the computer system, the MUDS having a bitmap field having a bit corresponding to each block of allocatable memory. A bit corresponding to a selected memory block is set to the value of &ldquo;1&rdquo; when the selected memory block is allocated to the selected process. The bit corresponding to the selected memory block is set to the value of &ldquo;0&rdquo; when the selected memory block is not allocated to the selected process. A master MUDS is generated by combining the MUDS maintained by each process, the master MUDS having bits set to a value of &ldquo;0&rdquo; for free memory blocks, and the master MUDS having bits set to a value of &ldquo;1&rdquo; for memory blocks allocated to any processes of the multiprocess computer system. In response to the master MUDS, all memory blocks having a corresponding bit set to a value of &ldquo;0&rdquo; are returned to free memory. Each process may execute on a different processor in a multiprocess or computer system, for example on interface processors of a router. In a router the memory usage data structure is referred to as the Buffer Usage Data Structure (BUDS). The master BUDS is generated, and any processor not submitting a processor BUDS does not have any bits in the master BUDS set to a value of &ldquo;1&rdquo;. Accordingly, any memory previously allocated to a processor which has crashed, or died, is then returned to the global free queue.

Distributed buffer recovery

A server installation, which includes multiple servers, services a client request using a filter index that is different than a destination address associated with the client request. This enables clients to generate client requests for a server installation in a conventional manner without regard to whether a server installation is formed by one server or multiple servers. Accordingly, when a server installation is scaled by increasing the number of servers for redundancy, load distribution or capacity reasons, reconfiguration of the clients utilizing the servers is unnecessary. In one arrangement, the data resides in a data structure having (i) a device identifier that uniquely identifies the server host among multiple server hosts, and (ii) a filter index which is different than the device identifier.

Methods and apparatus for transferring data using a filter index

The invention is directed to techniques for transferring data using a device driver that is arranged to prevent improper operation of a non-primary routine (e.g., an administrative operation) from causing improper operation of a primary routine (e.g., a data transfer operation). Accordingly, the primary routine can continue to operate properly after a failure of the non-primary routine. In one arrangement, a data communications device transfers data. The data communications device includes a port that couples to a network, and a processor coupled to the port. The data communications device further includes memory, coupled to the processor, that stores a device driver. The device driver has a first set of instructions that directs the processor to perform a data transfer routine that moves data between memory and the port, and a second set of instructions that directs the processor to perform an administrative routine. The second set of instructions is arranged to prevent improper operation of the administrative routine from causing improper operation of the data transfer routine. In particular, execution and communication mechanisms enable the primary and administrative routines to operate in such a fault-tolerant manner.

Methods and apparatus for transferring data using a device driver

A technique efficiently captures information required to create a filter used by switching nodes of a heterogeneous network to implement transmission priority routing of data traffic over a connection network between end nodes of the network. A bounded time interval, i.e., a time window, is established during which a first switching node monitors the traffic over the network to capture portions of the required information contained in a defined data packet. Monitoring of data traffic during the time window is triggered by a predictable event, the occurrence of which is communicated to the first switching node by a hybrid node of the network. In addition to providing this 'triggering' communication, the hybrid node also provides the first switching node with the remaining portion of the required information.

Technique for capturing information needed to implement transmission

A mechanism limits the flow of topology information to only those nodes of a computer network that require the information. The mechanism comprises a filter having a plurality of conditions that are evaluated by a node of the network prior to forwarding the topology information to another node. Specifically, the conditions are evaluated with respect to lists of nodes that are defined according to the configuration of the network. By applying the filtering conditions to the defined lists, each node may selectively propagate the topology information throughout the network, thereby substantially reducing the amount of information flowing over the network.

Technique for reducing the flow of topology information in a computer

A technique maintains priority of packets transmitted over a heterogeneous network by a hybrid node coupled to the network. The hybrid node assigns each packet a transmission priority (TP) level as it traverses protocol layers of a communications stack of the node. An application programming interface extension provides a tagging mechanism for conveying the TP levels from higher protocol layers of the stack to lower layers. The TP levels are then converted to information that is 'tagged' to the packets and these tagged packets are then apportioned among a plurality of communication sessions based on the TP level assigned to each packet. The tagged packets and their associated sessions preserve the priority and order of the packets across the heterogeneous network.

Growing community of inventors

Piedmont, CA, United States of America

Steven Howard Berl