Arista Networks, Inc.

Juniper Networks, Inc.

Kumaran Narayanan

Venkitraman Kasiviswanathan

Udayakumar Srinivasan

Abhijith Kudupu Narayan

Alton Lo

Sudip Regmi

Prashant Srinivas

Euan F Mowat

Surendra Patel

Venkata Duggirala

Peter Kwok

Saurabh Kumar Agarwal

Siva Madasamy

Sreedhar Ganjikunta

Wei Dai

Xiangyi Guo

Babu Neelam

Aparna Adhav

Kalash Nainwal

A method for generating an application-aware virtual topology (AAVT) routing table for a network device among network devices connected via a wide area network is provided. The method is executed by a network controller connected to the network and includes: receiving, from the network devices, path information of the network devices; generating, using the path information, an underlay graph specifying a path topology of the network device; generating, based on the path topology specified in the underlay graph, the AAVT routing table for the network device where the AAVT routing table includes a set of paths; and transmitting, in response to generating the AAVT routing table, the AAVT routing table to the network device to cause the network device to program the set of paths.

Systems and methods for constructing application-aware virtual topologies in wide area networks

A network element and method for programming a network element that includes detecting an update to a first route in a routing information base (RIB) is disclosed. The method includes locating a first route network prefix associated with the first route within a network prefix trie (NPT); determining that, prior to the update, a first parent network prefix and the first route network prefix were reachable using a pair of different next hops connected to the network element; and determining that, after the update, the first parent network prefix and the first route network prefix are reachable using a first common next hop connected to the network element. The method also includes removing an existing forwarding information base (FIB) entry in the FIB associated with the first route network prefix.

Efficient algorithm to eliminate redundant specific prefixes in forwarding information base using trie

In general, embodiments relates to a method for creating an on-demand tunnel (ODT) in a network between a first network device and a second network device, the method comprising: storing by the first network device, a potentially suboptimal path to the second network device, determining that a trigger condition to create the ODT between the first network device and the second network device is satisfied, in response to the determination: transmitting, by the first network device, an ODT signaling packet to the second network device via the potentially suboptimal path, receiving, from the second network device and in response to transmitting the ODT signaling packet, an ODT keepalive by first network device via the ODT, and transmitting, after receiving the ODT keepalive, a second packet to the second network device via the ODT.

On-demand setup and teardown of dynamic path selection tunnels

A method for transmitting network traffic across a wide area network (WAN) from a first site to a second site is provided. The method is executed by a first edge network device at the first site that further includes a second edge network device, and the method includes: receiving the network traffic from a client device at the first site; determining, using ipath characteristics and a classification of the network traffic, that the network traffic should be transmitted by the second edge network device to the second site; forwarding in response to the determination, the network traffic to the second edge network device using a local tunnel over a local area network (LAN) of the first site such that the network traffic is transmitted to the second site by the second edge network device.

Systems and methods for optimal routing in sites with multiple routers for high availability wide area networks

A method for servicing network traffic in a wide area network (WAN) comprising a plurality of network devices is provided. The method is executed by a network device among the plurality of network devices and comprises: receiving a request to transmit the network traffic to a destination network device where the request specifies that the network traffic is to be serviced by a network service; determining, based on the request and using a service-aware virtual topology (SAVT) routing table, a path through the WAN for reaching the network service and a service instance identifier (ID) of the network service; configuring the network traffic to include a service bit indicating whether service is to be performed and instructions specifying the path for reaching the network service; and transmitting, after configuring the network traffic, the network traffic toward the destination device through the at least one network service.

Systems and methods for constructing service-aware virtual topologies in wide area networks

Efficient algorithm to eliminate redundant specific prefixes in forwarding information base using TRIE

Systems and methods for load balancing in a network are disclosed. An illustrative method includes receiving network telemetry data corresponding to network paths of a plurality of coexisting multipaths, performing an adaptive load balancing process by determining whether a network path from the plurality of coexisting multipaths is an adequate network path based on the network telemetry data, and in response to determining the network path is an adequate network path, selecting the network path for a network flow.

Adaptive load balancing between routers in wan overlay networks using telemetry information

In general, the invention relates to a method for programming a network element. The method includes detecting an addition of a first route in a routing information base (RIB) on the network element, adding, in response to detecting the addition, a first route network prefix associated with the first route to a network prefix trie (NPT), identifying, based on the adding, a first parent network prefix for the first route network prefix using the NPT, making a first determination that the first route network prefix and the first parent network prefix are reachable via a first common next hop connected to the network element, and waiving, based on the first determination, a creation of a forwarding information base (FIB) entry associated with the first route network prefix in a FIB on the network element.

This disclosure describes a more efficient and configurable power allocation scheme for redundant power supply (RPS) systems used in network switches. This allocation scheme allows the system owner to assign power from a shared RPS unit to higher priority devices in any network switch in the system. This permits more granularity in assigning the RPS with backup power available to devices such as ports residing within individual switches in a multiple switch network. An efficient power allocation scheme for RPS allows the user to define the system priority of various devices for backup power according to the user's preferences. The user may assign the RPS to user-defined high priority devices in any piece of equipment. This makes RPS power allocation more flexible by offering the user more setup options for backup power.

Efficient power allocation for redundant power supply with system priority

Rolling software upgrades may be employed for a network device in a modular chassis and/or virtual chassis. The network device may include memory devices to store a software upgrade package and a group of instructions, and a processor. The processors may install the software upgrade package on a backup routing engine; determine subsets of multiple line cards on which to perform a software upgrade, where ports in each of the multiple line cards are part of a link aggregation group (LAG); initiate a reboot process for each of the subsets of multiple line cards, in sequence, where the reboot process for each of the line cards results in a software upgrade without deactivating any LAG. The processors may also switch the backup routing engine and a master routing engine to create a new master routing engine and a new backup routing engine, and install the upgrade package on the new backup routing engine.

Upgrading system software in a chassis without traffic loss

A method includes operating in a normal mode to receive and transmit packets, where the network device is one of multiple network devices that operate as a virtual chassis, where the virtual chassis corresponds to a single logical network device, and detecting when the network device crashes. The method further includes initiating a resetting process and operating in a pass through mode, during the resetting process, where the pass through mode permits packets to be received and transmitted to the network devices of the virtual chassis.

Growing community of inventors

San Ramon, CA, United States of America

Kumaran Narayanan