Browse Prior Art Database

NETWORK ANALYTICS TASK INTERACTIONS FRAMEWORK

IP.com Disclosure Number: IPCOM000240718D
Publication Date: 2015-Feb-20
Document File: 7 page(s) / 94K

Publishing Venue

The IP.com Prior Art Database

Related People

Alexander Clemm: AUTHOR [+3]

Abstract

A closed loop network query mechanism is provided by which the parameters of the initial query deployed at local nodes and/or applied to network telemetry of specific sources are subsequently modified based on global network conditions. This mechanism may be useful to collect fine-grained additional performance metrics only for a selected subset of network entities across a network. This functionality provides a real-time feedback mechanism to restrict attention to a subset of trending network entities.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 23% of the total text.

Page 01 of 7

NETWORK ANALYTICS TASK INTERACTIONS FRAMEWORK

AUTHORS:

 Alexander Clemm Mouli Chandramouli Sailesh Krishnamurthy

CISCO SYSTEMS, INC.

ABSTRACT

    A closed loop network query mechanism is provided by which the parameters of the initial query deployed at local nodes and/or applied to network telemetry of specific sources are subsequently modified based on global network conditions. This mechanism may be useful to collect fine-grained additional performance metrics only for a selected subset of network entities across a network. This functionality provides a real-time feedback mechanism to restrict attention to a subset of trending network entities.

DETAILED DESCRIPTION

     Analytics applications interact with network-generated data to obtain summarized information on the metrics that are of interest to network users. Distributed Network Analytics (DNA) proposes a query-based mechanism for obtaining the real-time network state from the network. In its simplest instantiation, the analytics query launched to network devices is fixed for the duration of the task and the results for the query are exported from the network devices. Thus, the query while running on a single network device does not have visibility of information from other neighboring nodes.

    An important category of network analytics use cases involves situations in which the precise device-level analytics that should be carried out at the local level depend on the results of other analytics carried out at a global level and network layer of abstraction. Examples involve cases in which detailed analytics required to "drill down" into some problem are required only at devices or interfaces that represent some kind of outlier at the global level. For example, a user may require detailed analysis of service level trends on the basis of precise service level measurements with a high sampling rate for devices

Copyright 2015 Cisco Systems, Inc.
1


Page 02 of 7

that show sustained uptrends in their packet drop rates that are network-wide in the top percentile. In this case, one part of the network analytics query (service level trend of measurements at a high sampling rate) is expensive but only needs to be performed if they match another part of the network analytics query (top percentile network-wide), for which part of the query term is not known by the device as it requires analysis of data from across the network.

    Presented herein are techniques that allow coupling of an analytics query obtaining network-level results with analytics queries performed at the device level in a seamless manner, making the local analytics adapt its query parameters as the analytics task progresses over time, resulting in self-adaptive properties based on global network conditions and results of network-wide analytics, involving closed-loop functionality between a network-level controller and networking devices. These techniques allow for continuous self-adaptation of analytics over time, with local nodes...