Data metric resolution ranking system and method
In one embodiment, a method includes identifying a plurality of period combinations for a metric of interest. Each period combination comprises one or more time periods and each of the one or more time periods comprises one or more segments. The method further includes, for each period combination of the plurality of period combinations and each historical value of a plurality of historical values of the metric of interest, incrementally inserting the historical value into corresponding segments of the one or more time periods. Moreover, the method includes ranking the plurality of period combinations based on comparisons of the plurality of historical values to sets of corresponding predicted values. The method also includes selecting a highest-ranked period combination as best characterizing the metric of interest.
1. A method comprising:
collecting over time, by a computer system comprising computer hardware, time-variable values of a performance metric of a computer resource;
identifying, by the computer system, a plurality of alternative period-combination data structures that each represent a same set of the collected time-variable values of the performance metric of the computer resource over a same period of time;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures:
the period-combination data structure comprises a different selection of two or more time periods of a plurality of time periods;
each time period of the two or more time periods comprises segments corresponding to divisions of said same period of time; and
the two or more time periods each span said same period of time at a different sampling rate, such that each of the two or more time periods includes a different number of segments;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures, the two or more time periods comprises a plurality of hierarchical time periods;
for each period-combination data structure of the plurality of alternative period-combination data structures and each time-variable value of the time-variable values of the performance metric, the computer system incrementally inserting the time-variable value into corresponding segments of the two or more time periods;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures, the incrementally inserting comprises, for each time period of the two or more time periods, iterating from highest-frequency time period to lowest-frequency time period:
identifying a corresponding segment of the time period for the time-variable value;
computing a predicted value for the corresponding segment;
adapting the time-variable value based on any higher-frequency time periods of the period-combination data structure;
wherein, if the time period is the highest-frequency time period of the two or more time periods, the adapted time-variable value comprises the time-variable value;
wherein, if the time period is not the highest-frequency time period of the two or more time periods, the adapted time-variable value comprises a difference between the adapted time-variable value and the predicted value for the corresponding segment of the immediately higher-frequency time period; and
inserting the adapted time-variable value for the time period into the corresponding segment of the time period;
ranking, by the computer system, the plurality of alternative period-combination data structures based on comparisons of the time-variable values to sets of corresponding predicted values;
selecting, by the computer system, a highest-ranked period-combination data structure as best characterizing the performance metric; and
the computer system predicting a future value of the performance metric for a specified future time point using the incrementally inserted time-variable values of the highest-ranked period-combination data structure;
the computer system alerting a user regarding the future value at the specified future time point; and
receiving a computer-system reconfiguration responsive to the alerting.
2. The method of claim 1 , comprising:
performing the identifying, the incrementally inserting, the ranking, and the selecting for each of a plurality of prediction algorithms;
assessing each highest ranked period-combination data structure; and
based on the assessment, selecting a particular prediction algorithm for the performance metric.
3. The method of claim 2 , wherein, for each prediction algorithm of the plurality of prediction algorithms, the computing of the predicted value comprises utilizing the prediction algorithm.
4. The method of claim 1 , wherein the ranking comprises, for each period-combination data structure of the plurality of alternative period-combination data structures:
comparing the time-variable values to a set of corresponding predicted values;
computing a score based on the comparison; and
ranking the period-combination data structure according to its score.
5. The method of claim 1 , comprising aggregating new time-variable values of the performance metric into corresponding segments of the highest-ranked period-combination data structure.
6. The method of claim 1 , wherein the ranking comprises, for each period-combination data structure of the plurality of alternative period-combination data structures:
calculating a linear correlation of a set of predicted values to a set of historical values;
computing a prediction interval;
performing prediction interval validation;
computing an outside area;
determining a score for the period-combination data structure based on the linear correlation, the prediction interval validation, and the outside area; and
ranking the period-combination data structure according to its score.
7. The method of claim 1 , comprising performing noise filtering of values of the time-variable values.
8. The method of claim 1 , comprising, for at least one time period of the highest-ranked period-combination data structure, performing noise filtering of the segments.
9. The method of claim 8 , wherein the performing of noise filtering of the segments comprises, for each segment of the segments:
dividing the segment into subsegments;
computing a segment noise value for each potential pairing of one of the subsegments with another of the subsegments;
selecting a maximum segment noise value; and
responsive to the maximum segment noise value exceeding a threshold, filtering the segment from the at least one time period.
10. An information handling system comprising at least one computer processor, wherein the at least one computer processor is operable to implement a method comprising:
identifying a plurality of alternative period-combination data structures that each represent a same set of the collected time-variable values of the performance metric of the computer resource over a same period of time;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures:
the period-combination data structure comprises a different selection of two or more time periods of a plurality of time periods;
each time period of the two or more time periods comprises segments corresponding to divisions of said same period of time; and
the two or more time periods each span said same period of time at a different sampling rate, such that each of the two or more time periods includes a different number of segments;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures, the two or more time periods comprises a plurality of hierarchical time periods;
for each period-combination data structure of the plurality of alternative period-combination data structures and each time-variable value of the time-variable values of the performance metric, incrementally inserting the time-variable value into corresponding segments of the two or more time periods;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures, the incrementally inserting comprises, for each time period of the two or more time periods, iterating from highest-frequency time period to lowest-frequency time period:
identifying a corresponding segment of the time period for the time-variable value;
computing a predicted value for the corresponding segment;
adapting the time-variable value based on any higher-frequency time periods of the period-combination data structure;
wherein, if the time period is the highest-frequency time period of the two or more time periods, the adapted time-variable value comprises the time-variable value;
wherein, if the time period is not the highest-frequency time period of the two or more time periods, the adapted time-variable value comprises a difference between the adapted time-variable value and the predicted value for the corresponding segment of the immediately higher-frequency time period; and
inserting the adapted time-variable value for the time period into the corresponding segment of the time period;
ranking the plurality of alternative period-combination data structures based on comparisons of the time-variable values to sets of corresponding predicted values;
selecting a highest-ranked period-combination data structure as best characterizing the performance metric; and
predicting a future value of the performance metric for a specified future time point using the incrementally inserted time-variable values of the highest-ranked period-combination data structure;
alerting a user regarding the future value at the specified future time point; and
receiving a computer-system reconfiguration responsive to the alerting.
11. The information handling system of claim 10 , comprising:
performing the identifying, the incrementally inserting, the ranking, and the selecting for each of a plurality of prediction algorithms;
assessing each highest ranked period-combination data structure; and
based on the assessment, selecting a particular prediction algorithm for the performance metric.
12. The information handling system of claim 10 , wherein the adapted time-variable value of a highest-frequency period of the plurality of hierarchical time periods comprises the time-variable value.
13. The information handling system of claim 10 , wherein the ranking comprises, for each period-combination data structure of the plurality of alternative period-combination data structures:
comparing the time-variable values to a set of corresponding predicted values;
computing a score based on the comparison; and
ranking the period-combination data structure according to its score.
14. A computer-program product comprising a non-transitory computer-usable medium having computer-readable program code embodied therein, the computer-readable program code adapted to be executed to implement a method comprising:
identifying a plurality of alternative period-combination data structures that each represent a same set of the collected time-variable values of the performance metric of the computer resource over a same period of time;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures:
the period-combination data structure comprises a different selection of two or more time periods of a plurality of time periods;
each time period of the two or more time periods comprises segments corresponding to divisions of said same period of time; and
the two or more time periods each span said same period of time at a different sampling rate, such that each of the two or more time periods includes a different number of segments;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures, the two or more time periods comprises a plurality of hierarchical time periods;
for each period-combination data structure of the plurality of alternative period-combination data structures and each time-variable value of the time-variable values of the performance metric, incrementally inserting the time-variable value into corresponding segments of the two or more time periods;
wherein, for each period-combination data structure of the plurality of alternative period-combination data structures, the incrementally inserting comprises, for each time period of the two or more time periods, iterating from highest-frequency time period to lowest-frequency time period:
identifying a corresponding segment of the time period for the time-variable value;
computing a predicted value for the corresponding segment;
adapting the time-variable value based on any higher-frequency time periods of the period-combination data structure;
wherein, if the time period is the highest-frequency time period of the two or more time periods, the adapted time-variable value comprises the time-variable value;
wherein, if the time period is not the highest-frequency time period of the two or more time periods, the adapted time-variable value comprises a difference between the adapted time-variable value and the predicted value for the corresponding segment of the immediately higher-frequency time period; and
inserting the adapted time-variable value for the time period into the corresponding segment of the time period;
ranking the plurality of alternative period-combination data structures based on comparisons of the time-variable values to sets of corresponding predicted values;
selecting a highest-ranked period-combination data structure as best characterizing the performance metric; and
predicting a future value of the performance metric for a specified future time point using the incrementally inserted time-variable values of the highest-ranked period-combination data structure;
alerting a user regarding the future value at the specified future time point; and
receiving a computer-system reconfiguration responsive to the alerting.