Adaptive unified performance management (AUPM) with root cause and/or severity analysis for broadband wireless access networks

• US 9,246,747 B2
• Filed: 11/15/2012
• Issued: 01/26/2016
• Est. Priority Date: 11/15/2012
• Status: Active Grant

• Current Assignee: Hong Kong Applied Science And Technology Research Institute Co., Ltd.
• Original Assignee: Hong Kong Applied Science And Technology Research Institute Co., Ltd.
• Inventors: Chan, Siu Ping, Mui, Ka Ho, Wang, Xiaodong, Chan, Chun Sing Joseph, Xiao, Shenglong, Xia, Hua, Liao, Qinghua
• Primary Examiner(s): Bates, Kevin
• Assistant Examiner(s): Mahmud, Golam
• Application Number:
  US13/678,362
• Publication Number:
  US 20140136685A1
• Time in Patent Office: 1,167 Days
• Field of Search: 709/224, 709/340, 709/370, 709/702, 709/714
• US Class Current: 1/1
• CPC Class Codes: H04L 41/022 Multivendor or multi-standard integration
• CPC Class Codes: H04L 41/0636 based on a decision tree analysis
• CPC Class Codes: H04L 41/142 using statistical or mathematical methods

Abstract

Systems and methods which provide an adaptive unified performance management (AUPM) framework for interacting with disparate network elements using techniques adaptive to operational conditions to provide network performance adaptive root cause analysis (ARCA) are shown. An AUPM framework of embodiments of the invention implements a proxy based architecture in which a plurality of proxies are utilized to connect to and perform data communication with the disparate network elements. Centralized performance management is in communication with the proxies to obtain and unify network element data for performance monitoring, alarm reporting, and/or root cause analysis. The performance monitoring, alarm reporting, and root cause analysis provided by centralized performance management of embodiments herein implements adaptive cluster-based analysis to provide robust operation adapted to accommodate various operational scenarios, such as may include time varying conditions and learning based configuration.

Claims

1. A method comprising:

obtaining a plurality of performance metrics from a plurality of network elements of monitored network communication infrastructure of a broadband wireless access network, wherein the obtaining the plurality of performance metrics includes utilizing multiple network equipment (MNE) data extractor logic collecting network element-specific data using a plurality of different data models for different network elements of the plurality of network elements and generating a common data model from the network element-specific data of the plurality of different data models for a particular network-wide performance aspect to be monitored, wherein the generating the common data model includes transforming the network element-specific data into a network-wide generic representation adapted to support monitoring of the network-wide performance aspect of the monitored network communication infrastructure;

generating a feature vector using the common data model, wherein the feature vector represents the network-wide performance aspect of the network communication infrastructure to be monitored; and

identifying at least one of a severity or a root cause of the network-wide performance aspect based on a distance of the feature vector to a centroid of a cluster of a plurality of clusters, wherein the cluster represents a root cause of the network-wide performance aspect to be monitored, and wherein the centroid of the cluster is updated using the feature vector generated from the common data model after the identifying.

2. The method of claim 1, further comprising:

generating the plurality of different data models specific to particular network elements of the plurality of network elements using management information bases (MIBs) and network element metadata.

3. The method of claim 2, wherein the MIBs define parameters that can be obtained from a corresponding network element of the plurality of network elements and the network element metadata includes data to characterize MIB parameters to be obtained from the corresponding network element for performance monitoring.

4. The method of claim 2, wherein the plurality of network elements include disparate network elements provided by different manufacturers and having different or proprietary management data interfaces, and wherein data models of the plurality of different data models are associated with network element groupings between the disparate network elements.

5. The method of claim 1, wherein one or more clusters of the plurality of clusters are associated with different performance aspect severities, and wherein a severity of the network-wide performance aspect to be monitored is identified as the severity associated with the cluster to which the feature vector is assigned.

6. The method of claim 1, further comprising:

updating the centroid of the cluster to which the feature vector is assigned as a function of the feature vector.

7. The method of claim 1, further comprising:

initializing the plurality of clusters using training data, wherein the initializing the plurality of clusters includes determining a centroid for each cluster of the plurality of clusters from a feature vector of the training data and iteratively refining the centroid using other feature vectors of the training data; and

updating the centroid for each cluster of the plurality of clusters to which a feature vector of the obtained plurality of performance metrics is associated as a function of that feature vector.

8. A system comprising:

a multiple network element data extractor adapted to communicate with disparate network elements of monitored network communication infrastructure of a broadband wireless access network and obtain performance metrics therefrom using a plurality of different data models corresponding to the disparate network elements, the multiple network element data extractor being further adapted to provide at least one common data model from the performance metrics of the plurality of different data models, wherein network element-specific data of the disparate network element performance metrics is transformed into a network-wide representation, homogenized as between the plurality of different data models, adapted to support monitoring of monitored network-wide performance aspects of the monitored network communication infrastructure;

performance monitoring logic adapted to obtain performance metrics from a common data model of the at least one common data model and provide feature vectors representative of the monitored network-wide performance aspects of the monitored network communication infrastructure;

adaptive root cause analysis logic adapted to implement adaptive cluster-based analysis of feature vectors of the feature vectors representative of the monitored network-wide performance aspects and identify a root cause of the monitored network-wide performance aspects based on a distance of the feature vectors to a centroid of a cluster of a plurality of clusters, wherein the cluster represents a root cause of a network-wide performance aspect of the monitored network-wide performance aspects, and wherein the centroid of the cluster is updated using the feature vector generated from the common data model after the root cause is identified; and

alarm reporting logic adapted to implement adaptive cluster-based analysis of feature vectors of the feature vectors representative of the monitored network-wide performance aspects and determine a severity of the monitored network-wide performance aspects for use in alarm reporting.

9. The system of claim 8, wherein the multiple network element data extractor comprises a data collection engine in communication with network elements of the monitored network communication infrastructure.

10. The system of claim 9, wherein the data collection engine is operable in accordance with a data collection profile of the multiple network element data extractor.

11. The system of claim 10, wherein the multiple network element data extractor further comprises a code and database generator, and wherein the data collection profile is generated by the code and database generator using management information bases and network element metadata associated with network elements of the disparate network elements.

12. The system of claim 11, wherein data models of the plurality of different data models are generated by the code and database generator using the management information bases and network element metadata associated with network elements of the disparate network elements.

13. The system of claim 9, further comprising:

a proxy based architecture including one or more proxy servers in communication with the disparate network elements and a centralized performance management server in communication with the one or more proxy servers, wherein at least a portion of the multiple network element data extractor including the data collection engine is operable upon the one or more proxy servers, and the performance monitoring logic and adaptive root cause analysis logic is operable upon the centralized performance management server.

14. The system of claim 8, wherein the disparate network elements are network elements provided by different manufacturers.

15. The system of claim 8, wherein the disparate network elements are network elements having different or proprietary management data interfaces.

16. A method comprising:

collecting network element-specific data from disparate network elements of monitored network communication infrastructure using a plurality of different data models, wherein data models of the plurality of different data models are associated with network element groupings between the disparate network elements;

generating a common data model from the network element-specific data of the plurality of different data models, wherein the generating the common data model includes transforming the network element-specific data into a network-wide generic representation adapted to support monitoring of a network-wide performance aspect of the monitored network communication infrastructure;

generating a feature vector using the common data model, wherein the feature vector represents the monitored network-wide communication performance aspect of the network infrastructure;

calculating a distance between the feature vector and a centroid of each cluster of a plurality of clusters;

assigning the feature vector to a cluster of the plurality of clusters having the centroid most near to the feature vector; and

identifying at least one of a root cause of the monitored network-wide performance aspect or a severity of the monitored network-wide performance aspect from the cluster to which the feature vector is assigned, wherein the cluster represents a root cause of the network-wide performance aspect, and wherein the centroid of the cluster is updated using the feature vector generated from the common data model after the identifying.

17. The method of claim 16, further comprising:

initializing the plurality of clusters using training data, wherein the initializing the plurality of clusters includes determining the centroid for each cluster of the plurality of clusters from a feature vector of the training data and iteratively refining the centroid using other feature vectors of the training data.