Cisco Technology, Inc.

Ankan Ghosh

Nataraj Bacthu

Apurup Reddy Chevuru

Swaminathan Narayanan

Laxmi Kant Agrawal

Ayan Banerjee

Harsha Bharadwaj

Raghava Kodigenahalli Sivaramu

Kishan Kumar Kotari Uppunda

Nataraj Batchu

Sunil John Varghese

Prabesh Babu Nanjundaiah

Sumanth Sindiri

The present disclosure involves systems and methods for managing a trie routing table for a networking device of a communication or computer network. In one implementation, the networking device may utilize a dynamic algorithm for associating hashing functions with pivot tiles of the routing table to improve hash utilization and avoid hash collisions. Further, route prefixes may be relocated from pivot tiles in an attempt to free the tiles for reallocation to other prefix base width or may be relocated to other possible pivot tiles or to a general storage space when a hash collision is detected. This provides for even distribution of pivots within tiles which have base widths in range of a pivot route. The above implementations may occur together or separately to improve the operation of the networking device and provide faster route lookup.

Dynamic route profile storage in a hardware trie routing table

Customized feature vectors are used to train a machine learning algorithm to automatically identify a network component where a network fault has occurred. A database comprising network components and associated network faults is analyzed to select a set of network components associated with the largest quantity of network faults. Customized features associated with the network faults are identified and selected for use in a feature vector as input to a machine learning algorithm. The features are selected based upon analysis of consistency checks, component configuration limits, and network wide configurations.

Identifying locations and causes of network faults

A computer-implemented method includes: obtaining platform dependent data and platform independent data from each of a plurality of first network devices for a computer network environment. The received platform dependent data and platform independent data are converted to key-value pairs. Sequences of the key-value pairs representing the configurations and states of the respective first network device are created. Vectors for the sequences of the key-value pairs are generated in a vector space. The vectors are supplied to a model to train the model to learn relationships of the platform dependent data and platform independent data of each of the plurality of first network devices. Platform dependent data and platform independent data from one or more second network devices are supplied to the trained software model to determine consistency of configurations and states of the one or more second network devices.

Machine learning method to validate run-time rules and detect software/hardware consistency errors for network elements

In one embodiment, a method includes receiving a request to add a prefix to memory for a route lookup at a forwarding device, the memory comprising a plurality of pivot tiles for storing pivot entries, each of the pivot entries comprising a plurality of prefixes and a pointer to a trie index, searching at the forwarding device, a dynamic pool of the pivot tiles based on a base-width associated with the prefix, allocating at least a portion of the pivot tile to the base-width and creating a pivot entry for the prefix and other prefixes with a corresponding base-width, and dynamically updating prefixes stored on the pivot tiles based on route changes to optimize storage of prefixes on the pivot tiles. An apparatus and logic are also disclosed herein.

Dynamic and compressed trie for use in route lookup

In one example embodiment, a time unconstrained fabric switch software upgrade function is described for an in-switch-software-upgrade/downgrade (ISSU) in a network switch. Prior to an ISSU, Hello protocol data unit (PDU) information is stored for relevant control protocols of associated interfaces of the switch. The Hello PDU information comprises information on messages and data exchanged by the control protocols that may be used to maintain communications links of the switch. During the configuring of the ISSU, and while the control plane of the switch is, at least partially, dysfunctional, Hello PDUs are sent on the interfaces according to the stored PDU information in order to maintain the communication links. The Hello PDU information may include a time interval value for each relevant protocol that determines when Hello PDUs are sent.

Unconstrained supervisor switch upgrade

Techniques are provided for bifurcating database information that might otherwise be replicated on each switch in a switched fabric of a Storage Area Network (SAN). The database is divided into a control plane database that comprises mostly switch specific data and a central management database that comprises user device configurations and device profiles. The control plane database includes information such as name server and zone server information, for those devices that may locally log into the switched fabric via a given switch, and those remote devices that are zoned with those local devices and that may log into the switched fabric via another switch. The central management database includes global information for the switched fabric and device profile information (e.g., login interface, Virtual SAN membership, device aliasing, etc.) for devices that have access to the switched fabric and that can also be requested by the various switches in the switched fabric.

Growing community of inventors

San Jose, CA, United States of America

Ankan Ghosh