System and method for detecting system relationships by correlating system workload activity levels
Relationships between systems can be inferred through a correlation analysis of the system workload activity levels. A method, computer readable medium and system are provided for analyzing correlations between the system workloads. The method comprises obtaining a set of quantile-based workload data pertaining to a plurality of systems. The correlation coefficient limit may then be used to compute the workload correlation scores for the plurality of systems and a result indicative of relationships between the systems then provided.
1. A method for detecting relationships between systems based on correlations between system workloads, said method comprising:
obtaining a set of quantile-based workload data pertaining to a plurality of systems;
converting said quantile-based workload data to relative measures if appropriate;
computing a correlation coefficient of workloads for a plurality of relevant system combinations;
computing a correlation score using a correlation coefficient for each system combination; and
providing analysis results in a correlation map.
2. The method according to claim 1 further comprising obtaining a set of time series workload data and converting said time series workload data into said quantile-based workload data.
3. The method according to claim 2 wherein said time series data is collected from said plurality of systems using data collection inputs.
4. The method according to claim 2 wherein said quantile-based workload data is converted according to quantile and time basis inputs.
5. The method according to claim 1 wherein said quantile-based workload data is selected for said plurality of systems from a stored set of quantile-based workload data according to workload data selection criteria inputs.
6. The method according to claim 1 further comprising determining if a relative correlation analysis is applicable, said converting comprising dividing values in said quantile-based workload data by an average value.
7. The method according to claim 1 wherein said correlation coefficient C, is computed as a normalization of a standard deviation as follows:
C
=
100
N
*
x
_
∑
n
N
(
x
n
-
x
_
)
2
;
where x n is the n th data value instance; x is the average of the data values; and N is the number of data values.
8. The method according to claim 1 further comprising computing an overall correlation coefficient for multiple time samples as follows:
C
=
100
T
*
N
*
x
_
∑
t
T
∑
n
N
(
x
t
,
n
-
x
_
t
)
2
;
where: C is the correlation coefficient for the time period; N is the number of systems being analyzed; n is the n th system; x t,n is the data sample for the n th system at time sample t; x t is the average data value for all the systems at time sample t; x is the overall average value for all the systems for all the time samples; T is the total number of time samples analyzed; and t is the t th time series data sample (e.g. 0, 1, 2, 3, . . . , 23).
9. The method according to claim 1 wherein correlation coefficients for multiple systems are computed for a set of statistics and multiple time sample such that for each time sample, the coefficients for multiple statistics are combined as a weighted average.
10. The method according to claim 9 wherein the correlation coefficient for a set of systems for multiple times samples and multiple statistics is computed as follows:
C
=
100
T
*
N
*
x
_
∑
t
T
∑
s
S
[
w
s
*
∑
n
N
(
x
t
,
s
,
n
-
x
_
t
,
s
)
2
]
;
where: C is the correlation coefficient; N is the number of systems being analyzed; T is the total number of time samples; S is the total number of statistical data values; x t,s,n is the data sample for the n th system for statistic s at time sample t; x t,s is the average data value for the N systems for statistic s at time sample t; x is the overall average value for all the systems for all the statistics and time samples; w s is the weighting factor for combining the statistics; n is the n th system; s is the s th statistic data value; and t is the t th time series data sample.
11. The method according to claim 1 wherein said correlation scores are computed using a correlation coefficient limit.
12. The method according to claim 8 wherein said correlation scores are computed as follows:
S
=
100
*
(
1
-
C
-
L
L
)
;
where: S is said correlation score; C is said correlation coefficient; and L is said coefficient limit.
13. The method according to claim 1 , further comprising sorting said systems based on said correlation scores for organizing said correlation map.
14. The method according to claim 1 wherein results in said correlation map are presented in a grid wherein systems are listed as rows and columns and each cell in said grid comprise a number corresponding to a correlation score and is color coded to provide a visual indication of the extent of correlation and to facilitate sorting said correlation scores in said correlation map.
15. A computer readable medium comprising computer executable instructions for performing the method according to claim 1 .
16. A system for estimating combined system workloads, said system comprising a database for storing workload data, said computer readable medium according to claim 15 , and a processor for executing said computer executable instructions.