SYSTEM AND METHOD FOR ROOT CAUSE ANALYSIS OF MOBILE NETWORK PERFORMANCE PROBLEMS
A method and system for identifying causes of performance metric changes in a network by selecting, from a pool of network event counters, a plurality of candidate counters relevant to a performance metric; grouping the candidate counters into clusters of similar counters; selecting, from each cluster, one or more representative counters; and fitting the selected representative counters to a model of the performance metric to determine thereby a set of representative counters most relevant to the performance metric.
1 . A method for identifying causes of performance metric changes in a network, the method comprising:
selecting, from a pool of network event counters, a plurality of candidate counters relevant to a performance metric;
grouping candidate counters into clusters of similar counters;
selecting, from each cluster, one or more representative counters; and
fitting the selected representative counters to a model of the performance metric to determine thereby a set of representative counters most relevant to the performance metric.
2 . The method of claim 1 , further comprising normalizing the values of said selected plurality of candidate counters.
3 . The method of claim 1 , wherein selecting the plurality of candidate counters comprises:
determining for each event counter a respective importance score for the performance metric; and
selecting as candidate counters for the performance metric those event counters having a respective importance score above a threshold level.
4 . The method of claim 3 , wherein said importance score is determined according to a rank correlation.
5 . The method of claim 4 , wherein said rank correlation comprises a Pearson correlation.
6 . The method of claim 3 , wherein said importance score is determined according to a Komogorov-Smirnov (KS) test.
7 . The method of claim 1 , wherein grouping the candidate counters into the clusters comprises:
computing a correlation between pairs of candidate counters to provide a plurality of nodes, wherein node edges are defined when an absolute value of a respective correlation exceeds a threshold level.
8 . The method of claim 1 , wherein said one or more representative counters comprises a single candidate counter having the largest correlation to the performance metric.
9 . The method of claim 1 , wherein said one or more representative counters comprises a predefined number of candidate counters having the largest correlation to the performance metric.
10 . The method of claim 1 , wherein said one or more representative counters comprises a set of candidate counters having a correlation to the performance metric above a threshold level.
11 . The method of claim 1 , wherein said fitting uses a regression analysis.
12 . The method of claim 1 , wherein said fitting uses a classification tree.
13 . The method of claim 1 , wherein said fitting uses a classification/regression tree adapted in accordance with a boosting procedure.
14 . The method of claim 13 , wherein said boosting procedure comprises an AdaBoost method.
15 . The method of claim 1 , wherein of said method is repeated for each of a plurality of performance metrics.
16 . The method of claim 1 , wherein said grouping is performed using one or more statistical clustering techniques.
17 . The method of claim 16 , wherein said statistical clustering techniques comprise any of a spectral clustering technique, a hierarchical clustering technique and a cost tree analysis technique.
18 . An apparatus for use in a network management system and for identifying causes of performance metric changes in a network, the apparatus comprising:
a processor configured to:
select, from a pool of network event counters, a plurality of candidate counters relevant to a performance metric;
group candidate counters into clusters of similar counters;
select, from each cluster, one or more representative counters; and
fit the selected representative counters to a model of the performance metric to determine thereby a set of representative counters most relevant to the performance metric.
19 . A tangible and non-transitory computer readable medium including software instructions which, when executed by a processer, perform a method for identifying causes of performance metric changes in a network, the method comprising:
selecting, from a pool of network event counters, a plurality of candidate counters relevant to a performance metric;
grouping candidate counters into clusters of similar counters;
selecting, from each cluster, one or more representative counters; and
fitting the selected representative counters to a model of the performance metric to determine thereby a set of representative counters most relevant to the performance metric.
20 . A computer program product, wherein computer instructions, when executed by a processor in a computer, perform a method for identifying causes of performance metric changes in a network, the method comprising:
selecting, from a pool of network event counters, a plurality of candidate counters relevant to a performance metric;
grouping candidate counters into clusters of similar counters;
selecting, from each cluster, one or more representative counters; and
fitting the selected representative counters to a model of the performance metric to determine thereby a set of representative counters most relevant to the performance metric.