Browse Prior Art Database

INTELLIGENT NETWORK COMPLIANCE

IP.com Disclosure Number: IPCOM000248357D
Publication Date: 2016-Nov-17
Document File: 4 page(s) / 375K

Publishing Venue

The IP.com Prior Art Database

Related People

Vinit Jain: AUTHOR [+2]

Abstract

Described herein are methods for using machine learning techniques, such as supervised learning, to provide network compliance recommendations by training the machine learning system with production environment data.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 4

INTELLIGENT NETWORK COMPLIANCE

AUTHORS:

   Vinit Jain
Gonzalo Salgueiro

CISCO SYSTEMS, INC.

ABSTRACT

    Described herein are methods for using machine learning techniques, such as supervised learning, to provide network compliance recommendations by training the machine learning system with production environment data.

DETAILED DESCRIPTION

     Different customers run various features, software, hardware and configurations on equipment in their networks. However, different combinations of these elements can lead to different types of defects that users face in their network. Users are likely run into these defects due to their use of different combinations of hardware, software, features and configurations.

    Methods described herein involve leveraging a variety of Technical Assistance Center (TAC)/customer data sources (including Customer Relationship Management (CRM)/case notes, product telemetry, device configurations, log files, debugs, bugs, etc.) and data from collectors deployed at various customer sites to define, via a machine- learning (ML) algorithm, various network compliance characteristics for different customers based on their feature usage and the scale of their network.

    These methods involve proactively feeding the data to the ML system with different combinations of inputs from various customers, including software version, hardware type, features, configuration, and scale.

    Customer network information related to the hardware and software configuration, feature configuration, features running on the network nodes, and scale of the network can be all fetched from tools such as collectors / real time telemetry data mechanisms and stored in the backend database.

Copyright 2016 Cisco Systems, Inc.

1


Page 02 of 4

    In addition to the datasets, another set of data is collected from customers when the customers open TAC cases for issues involving different features, configuration, and scale, including issues involving a defect related to the feature for which the case was opened. This data is also saved into the backend as another dataset.

    Both the datasets are fed into the ML system, as illustrated in Figure 1 below. Using ML techniques such as supervised learning, the system can learn about the stability of different customer environments for a feature running on particular hardware and software versions along with the comb...