Methods for monitoring computer resources using a first and second matrix, and a feature relationship tree

• US 8,719,196 B2
• Filed: 12/19/2011
• Issued: 05/06/2014
• Est. Priority Date: 12/19/2011
• Status: Active Grant

• Current Assignee: Go Daddy Operating Company LLC (GoDaddy, Inc.)
• Original Assignee: Go Daddy Operating Company LLC (GoDaddy, Inc.)
• Inventors: Thompson, Erskine
• Primary Examiner(s): VINCENT, DAVID ROBERT
• Application Number:
  US13/330,346
• Publication Number:
  US 20130159502A1
• Time in Patent Office: 869 Days
• Field of Search: 706/12, 706/15, 706/62, 709224-225, 718/100, 718104- 5
• US Class Current: 706/15
• CPC Class Codes: G06F 11/0709 in a distributed system consisting of a plurality of standalone computer nodes, e.g. clusters, client-server systems
• CPC Class Codes: G06F 11/0754 by exceeding limits
• CPC Class Codes: G06F 11/3409 for performance assessment
• CPC Class Codes: G06F 11/3447 Performance evaluation by modeling
• CPC Class Codes: G06F 2201/81 Threshold
• CPC Class Codes: H04L 63/0815 providing single-sign-on or federations

Abstract

One embodiment of a method of the present invention for monitoring computer resources provides for retrieving a set of resource-metric records for a predetermined time interval, forming a first mathematical matrix containing metric'"'"'s values arranged on date-time and resource-metric axes, creating a second mathematical matrix containing features and a third mathematical matrix containing weights, building a feature relationship tree, generating a predicted value for the resource-metric identifier, determining a variance between predicted value and metric'"'"'s value, and triggering an alert if the variance exceeds a predetermined alert threshold.

Claims

1. A method, comprising the steps of:

a) retrieving a first set of records for a first predetermined time interval from one or more resource-metric records stored at one or more electronic storage devices, wherein said one or more resource-metric records at least contain a resource-metric identifier, a metric'"'"'s value, and a date-time value when said metric'"'"'s value was obtained,

b) forming a first mathematical matrix containing one or more metric'"'"'s values arranged on date-time and resource-metric axes,

c) creating a second mathematical matrix containing one or more features and a third mathematical matrix containing one or more weights, wherein multiplication of said second mathematical matrix and said third mathematical matrix produces said first mathematical matrix,

d) building a feature relationship tree, wherein each feature in said one or more features is assigned a list of date-time values and date-time weights from said third mathematical matrix, and said each feature is assigned a list of resource-metric identifiers and resource-metric weights from said second mathematical matrix,

e) generating a predicted value for said resource-metric identifier and a most recent date-time value in said feature relationship tree,

f) determining a variance between said predicted value and said metric'"'"'s value for said resource-metric identifier, and

g) if said variance exceeds a predetermined alert threshold, then triggering an alert.

2. The method of claim 1, further comprising the step of:

h) collecting said one or more resource-metric records from one or more resources on a computer network.

3. The method of claim 1, further comprising the step of:

h) saving said one or more resource-metric records to said one or more electronic storage devices.

4. The method of claim 1, wherein said creating said second mathematical matrix and said third mathematical matrix is achieved by applying a non-negative matrix factorization algorithm to said first mathematical matrix.

5. The method of claim 1, further comprising the step of:

h) sorting said list of date-time values and date-time weights for said each feature by date-time weight in descending order in said feature relationship tree.

6. The method of claim 1, further comprising the step of:

h) sorting said list of resource-metric identifiers and resource-metric weights for said each feature by resource-metric weight in descending order in said feature relationship tree.

7. The method of claim 1, further comprising the step of:

h) sorting said features by highest date-time weight of said each feature in descending order in said feature relationship tree.

8. The method of claim 1, further comprising the step of:

h) pruning nodes in said list of date-time values and date-time weights by leaving top N1 nodes by date-time weights in said each feature and deleting nodes with sequential positions more than N1, wherein N1 is a first predetermined value.

9. The method of claim 1, further comprising the step of:

h) pruning nodes in said list of resource-metric identifiers and resource-metric weights by leaving top N2 nodes by resource-metric weights in said each feature and deleting nodes with sequential positions more than N2, wherein N2 is a second predetermined value.

10. The method of claim 1, further comprising the step of:

h) deleting features from said feature relationship tree if said features do not contain a node corresponding to said most recent date-time value in said first set of records retrieved from said one or more electronic storage devices.

11. The method of claim 1, wherein said generating said predicted value for said resource-metric identifier and said most recent date-time value in said feature relationship tree is achieved by averaging said metric'"'"'s values.

12. The method of claim 1, wherein said generating said predicted value for said resource-metric identifier and said most recent date-time value in said feature relationship tree is achieved via a k-Nearest Neighbor algorithm.

13. The method of claim 1, wherein said variance between said predicted value and said metric'"'"'s value is normalized on a sliding scale.

14. The method of claim 1, wherein said variance between said predicted value and said metric'"'"'s value is normalized on a scale between 0 and 1.

15. The method of claim 1, wherein said variance between said predicted value and said metric'"'"'s value is normalized using a Gaussian function.

16. The method of claim 1, further comprising the step of:

h) verifying that a latest metric'"'"'s value for said resource-metric identifiers will trigger said alert before triggering said alert.

17. A method, comprising the steps of:

a) collecting one or more resource-metric records from one or more resources on a computer network, wherein said one or more resource-metric records at least contain a resource-metric identifier, a metric'"'"'s value, and a date-time value when said metric'"'"'s value was obtained,

b) saving said one or more resource-metric records to one or more electronic storage devices,

c) retrieving a first set of records from said one or more resource-metric records for a first predetermined time interval from said one or more electronic storage devices,

d) forming a first mathematical matrix containing one or more metric'"'"'s values arranged on date-time and resource-metric axes,

e) creating a second mathematical matrix containing one or more features and a third mathematical matrix containing one or more weights by applying a non-negative matrix factorization algorithm to said first mathematical matrix, wherein multiplication of said second mathematical matrix and said third mathematical matrix produces said first mathematical matrix,

f) building a feature relationship tree, wherein each feature in said one or more features is assigned a list of date-time values and date-time weights from said third mathematical matrix, and said feature is assigned a list of resource-metric identifiers and resource-metric weights from said second mathematical matrix,

g) sorting said list of date-time values and date-time weights for said each feature by date-time weight in descending order in said feature relationship tree,

h) sorting said list of resource-metric identifiers and resource-metric weights for said each feature by resource-metric weight in descending order in said feature relationship tree,

i) sorting said features by highest date-time weight of said each feature in descending order in said feature relationship tree,

j) pruning nodes in said list of date-time values and date-time weights by leaving top N1 nodes by date-time weights in said each feature and deleting nodes with sequential positions more than N1, wherein N1 is a first predetermined value,

k) pruning nodes in said list of resource-metric identifiers and resource-metric weights by leaving top N2 nodes by resource-metric weights in said each feature and deleting nodes with sequential positions more than N2, wherein N2 is a second predetermined value,

l) deleting features from said feature relationship tree if said features do not contain a node corresponding to a most recent date-time value in said first set of records retrieved from said one or more electronic storage devices,

m) generating a predicted value for each resource-metric identifier and said most recent date-time value in said feature relationship tree via a k-Nearest Neighbor algorithm,

n) determining a variance between said predicted value and said metric'"'"'s value for said each resource-metric identifier as a normalized value between 0 and 1 using a Gaussian function for said each resource-metric identifier, and

o) if said variance exceeds a predetermined alert threshold, then triggering an alert.

18. A method, comprising the steps of:

a) retrieving a first set of records for a single resource and a first predetermined time interval from one or more metric records stored at one or more electronic storage devices, wherein said one or more metric records at least contain a metric identifier, a metric'"'"'s value, and a date-time value when said metric'"'"'s value was obtained,

b) forming a first mathematical matrix containing one or more metric'"'"'s values arranged on date-time and metric identifier axes,

c) creating a second mathematical matrix containing one or more features and a third mathematical matrix containing one or more weights, wherein multiplication of said second mathematical matrix and said third mathematical matrix produces said first mathematical matrix,

d) building a feature relationship tree, wherein each feature in said one or more features is assigned a list of date-time values and date-time weights from said third mathematical matrix, and said each feature is assigned a list of metric identifiers and metric identifier weights from said second mathematical matrix,

e) generating a predicted value for said metric identifier and a most recent date-time value in said feature relationship tree,

f) determining a variance between said predicted value and said metric'"'"'s value for said metric identifier, and

g) if said variance exceeds a predetermined alert threshold, then triggering an alert.

19. A method, comprising the steps of:

a) retrieving a first set of records for a single metric and a first predetermined time interval from one or more resource records stored at one or more electronic storage devices, wherein said one or more resource records at least contain a resource identifier, a metric'"'"'s value, and a date-time value when said metric'"'"'s value was obtained,

b) forming a first mathematical matrix containing one or more metric'"'"'s values arranged on date-time and resource identifier axes,

c) creating a second mathematical matrix containing one or more features and a third mathematical matrix containing one or more weights, wherein multiplication of said second mathematical matrix and said third mathematical matrix produces said first mathematical matrix,

d) building a feature relationship tree, wherein each feature in said one or more features is assigned a list of date-time values and date-time weights from said third mathematical matrix, and said each feature is assigned a list of resource identifiers and resource identifier weights from said second mathematical matrix,

e) generating a predicted value for said resource identifier and a most recent date-time value in said feature relationship tree,

f) determining a variance between said predicted value and said metric'"'"'s value for said resource identifier, and

g) if said variance exceeds a predetermined alert threshold, then triggering an alert.