Anomaly detection in data protection operations
Described herein are techniques for better understanding problems arising in an illustrative information management system, such as a data storage management system, and for issuing appropriate alerts and reporting to data management professionals. The illustrative embodiments include a number of features that detect and raise awareness of anomalies in system operations. Categories of interest include events and job anomalies, such as long-running jobs and job success/failure rates. Anomalies are characterized by frequency anomalies and/or by occurrence counts. Utilization is also of interest for certain key system resources, such as deduplication databases, CPU and memory at the storage manager, etc., without limitation. Predicting low utilization periods for these and other key resources is useful for scheduling maintenance activities without interfering with ordinary data protection jobs.
1. A networked information management system comprising:
a client computing device having one or more first hardware processors, wherein a first type of event occurs on the client computing device; and
one or more computing devices in communication with the client computing device, wherein the one or more computing devices are configured with computer-executable instructions that, when executed, cause the one or more computing devices to:
retrieve event data corresponding to the first type of event and the client computing device;
perform a time-series decomposition of the event data;
analyze a component of the decomposed time-series to determine an acceptable range for a number of occurrences of the first type of event;
determine not to expand the acceptable range in response to an indication that a number of alerts generated for the first type of event is less than a threshold;
determine that an anomaly exists at a first time in response to a determination that a number of occurrences of the first type of event falls outside the acceptable range; and
generate an alert for the detected anomaly.
2. The networked information management system of claim 1 , wherein the computer-executable instructions, when executed, further cause the one or more computing devices to perform the time-series decomposition of the event data to form a trend component, a seasonal component, and an error component.
3. The networked information management system of claim 2 , wherein the computer-executable instructions, when executed, further cause the one or more computing devices to analyze the error component to determine the acceptable range for the number of occurrences of the first type of event.
4. The networked information management system of claim 1 , wherein the computer-executable instructions, when executed, further cause the one or more computing devices to determine that a second anomaly exists at the first time in response to a determination that a duration between occurrences of the first type of event falls outside a second acceptable range.
5. The networked information management system of claim 4 , wherein the duration between occurrences of the first type of event is less than a lower extreme of the second acceptable range.
6. The networked information management system of claim 1 , wherein the number of occurrences of the first type of event is greater than an upper limit of the acceptable range.
7. The networked information management system of claim 1 , wherein the number of occurrences of the first type of event is less than a lower limit of the acceptable range.
8. A computer-implemented method comprising:
retrieving event data corresponding to a first type of event that occurs on a client computing device;
performing a time-series decomposition of the event data;
analyzing a component of the decomposed time-series to determine an acceptable range for a number of occurrences of the first type of event;
determining not to expand the acceptable range in response to an indication that a number of alerts generated for the first type of event is less than a threshold;
determining that an anomaly exists at a first time in response to a determination that a number of occurrences of the first type of event falls outside the acceptable range; and
generating an alert for the detected anomaly.
9. The computer-implemented method of claim 8 , wherein performing the time-series decomposition further comprises performing the time-series decomposition of the event data to form a trend component, a seasonal component, and an error component.
10. The computer-implemented method of claim 9 , wherein analyzing a component of the decomposed time-series further comprises analyzing the error component to determine the acceptable range for the number of occurrences of the first type of event.
11. The computer-implemented method of claim 8 , further comprising determining that a second anomaly exists at the first time in response to a determination that a duration between occurrences of the first type of event falls outside a second acceptable range.
12. The computer-implemented method of claim 11 , wherein the duration between occurrences of the first type of event is less than a lower extreme of the second acceptable range.
13. The computer-implemented method of claim 8 , wherein the number of occurrences of the first type of event is greater than an upper limit of the acceptable range.
14. The computer-implemented method of claim 8 , wherein the number of occurrences of the first type of event is less than a lower limit of the acceptable range.
15. A non-transitory computer-readable medium storing instructions, which when executed by one or more computing devices, cause the one or more computing devices to perform a method comprising:
retrieving event data corresponding to a first type of event that occurs on a client computing device;
performing a time-series decomposition of the event data;
analyzing a component of the decomposed time-series to determine an acceptable range for a number of occurrences of the first type of event;
determining not to expand the acceptable range in response to an indication that a number of alerts generated for the first type of event is less than a threshold;
determining that an anomaly exists at a first time in response to a determination that a number of occurrences of the first type of event falls outside the acceptable range; and
generating an alert for the detected anomaly.
16. The non-transitory computer-readable medium of claim 15 , wherein the method further comprises performing the time-series decomposition of the event data to form a trend component, a seasonal component, and an error component.
17. The non-transitory computer-readable medium of claim 16 , wherein the method further comprises analyzing the error component to determine the acceptable range for the number of occurrences of the first type of event.
18. The non-transitory computer-readable medium of claim 15 , wherein the method further comprises determining that a second anomaly exists at the first time in response to a determination that a duration between occurrences of the first type of event falls outside a second acceptable range.
19. The non-transitory computer-readable medium of claim 18 , wherein the duration between occurrences of the first type of event is less than a lower extreme of the second acceptable range.
20. The non-transitory computer-readable medium of claim 15 , wherein the number of occurrences of the first type of event is greater than an upper limit of the acceptable range.