Granted Patent B2

US 9,350,627 · App. 13/650,767 · Granted May 24, 2016

Abstract

Claims (49)

1. A method for upgrading or allocating resources, the method comprising:

collecting, by least at least one processor, a dataset by sampling utilization versus workload of resources in an Information Technology (IT) computer system, the resources of the IT computer system including hardware and software IT resources, the dataset including aggregate measurements of workload and utilization over time;

analyzing, by the least one processor, the collected dataset to obtain a service time estimation using a refinement enhanced clusterwise regression process, the service time estimation being a parameter relating to an estimated time required to process a request by the IT computer system, the refinement enhanced clusterwise regression process including:

normalizing, by the least one processor, the collected dataset;

scattering, by the least one processor, data from the normalized dataset;

obtaining, by the least one processor, a plurality of clusters based on the scattered data;

discarding, by the least one processor, one or more clusters from the plurality of clusters with less than a percentage of a total number of observations;

in each cluster, performing, by the least one processor, clusterwise regression and obtaining linear sub-clusters in a defined number;

reducing, by the least one processor, one or more sub-clusters including applying a refinement procedure, removing one or more sub-clusters that fit to outliers and merging pairs of clusters that fit an equivalent model;

updating, by the least one processor, one or more clusters with the reduced sub-clusters;

removing, by the least one processor, one or more globular clusters;

reducing, by the least one processor, a number of clusters with the refinement procedure; and

de-normalizing, by the least one processor, data of the reduced number of clusters, wherein the service time estimation is computed from a slope of a regression line of a given cluster in the reduced number of clusters; and

upgrading or allocating at least a portion of the resources of the IT computer system based on the service time estimation.

2. The method of claim 1 , wherein the refinement procedure includes a merging step, wherein a delta is computed for each cluster and then a pair that, when merged, gives origin to the cluster with a smallest increase in cluster delta is found, then

if the increase of the delta is below a threshold, the pair of clusters is merged and one or more new regression lines is computed and then the refinement procedure is started again, otherwise the merging step is ended.

3. The method of claim 2 , wherein the refinement procedure provides that,

given a cluster C 1 , an associated regression line defined by coefficients (R i , S f ), and a point (X j , U j ), let d(i,j) be the orthogonal distance of the point from the regression line, the following steps are performed,

for each cluster C i the distances d(i,j) for j=1, |C i | is computed and considered a random sample from an unknown distribution, and

assuming δρ(C i ) the p-percentile of said sample, a point j is considered inliner with respect to a cluster if d(i,j)<1.5δ 0.9 (C i ), then

if more than a certain percentage T i of the points of the cluster are inliners with respect to other clusters or if less than T p points are not inliners with respect to other clusters, the cluster is removed, its points reassigned to the closest cluster and a local search is performed.

4. The method of claim 2 , wherein the refinement procedure provides

assigning the points of one cluster to other clusters and checking which cluster suffers the biggest delta increase, then finding the cluster that, when having all its points assigned to other clusters, gives origin to the smallest max increase in delta, then

if delta increase is below a predetermined threshold, said cluster is actually removed.

5. The method of claim 1 , further comprising a visual mining procedure including:

calculating a silhouette value for each point to measure strength of point to cluster membership;

choosing a value for a threshold; and

outputting at least one chart.

6. The method of claim 1 , wherein the at least one chart includes a silhouette-time chart, a cardinality-time chart, a parameter-time chart, an hour of day chart, a day of week chart, and a timetable chart.

7. A non-transitory computer readable medium including executable instructions that when executed cause at least one processor to:

collect a dataset by sampling utilization versus workload of resources in an Information Technology (IT) computer system, the resources of the IT computer system including hardware and software IT resources, the dataset including aggregate measurements of workload and utilization over time;

analyze the collected dataset to obtain a service time estimation using a refinement enhanced clusterwise regression process, the service time estimation being a parameter relating to an estimated time required to process a request by the IT computer system, the refinement enhanced clusterwise regression process including:

normalize the collected dataset;

scatter data from the normalized dataset;

obtain a plurality of clusters based on the scattered data;

discard one or more clusters from the plurality of clusters with less than a percentage of a total number of observations;

in each cluster, perform clusterwise regression and obtaining linear sub-clusters in a defined number;

reduce one or more sub-clusters including applying a refinement procedure, removing one or more sub-clusters that fit to outliers and merging pairs of clusters that fit an equivalent model;

update one or more clusters with the reduced sub-clusters;

remove one or more globular clusters;

reduce a number of clusters with the refinement procedure; and

de-normalize data of the reduced number of clusters, wherein the service time estimation is computed from a slope of a regression line of a given cluster in the reduced number of clusters; and

upgrading or allocating at least a portion of the resources of the IT computer system based on the service time estimation.

8. The non-transitory computer readable medium of claim 7 , wherein the refinement procedure includes a merging step, wherein a delta is computed for each cluster and then a pair that, when merged, gives origin to the cluster with a smallest increase in cluster delta is found, then if the increase of the delta is below a threshold, the pair of clusters is merged and one or more new regression lines is computed and then the refinement procedure is started again, otherwise the merging step is ended.

9. The non-transitory computer readable medium of claim 7 , wherein the executable instructions cause the at least one processor to:

calculate a silhouette value for each point to measure strength of point to cluster membership;

choose a value for a threshold; and

output at least one chart.

10. The non-transitory computer readable medium of claim 9 , wherein the at least one chart includes a silhouette-time chart, a cardinality-time chart, a parameter-time chart, an hour of day chart, a day of week chart, and a timetable chart.

Assignments (15)