Storage mounting event failure prediction

1. A computer-implemented method for failure prediction of a queried mount event when mounting a queried storage medium by a queried media drive, the method comprising:

providing, by a processor, a machine learning model having an input and an output;

receiving input data, the input data including log data of the queried storage medium and the queried media drive, wherein the input data is generated by forcing a memory dump while the queried storage medium is mounted by the queried media drive, and by a memory readout while the storage medium is stored outside of the queried media drive;

providing the input data to the input of the machine learning model;

determining, from the output of the machine learning model, a predicted failure cause category and a predicted failure probability assigned to the predicted failure cause category; and

providing a first prediction to a user.

2. The method of claim 1 , wherein the predicted failure cause category is a category from a predefined set of failure cause categories, the predefined set of failure cause categories including a medium category associated with the queried storage medium, a drive category associated with the queried media drive, and a permit category associated with the absence of a failure cause, wherein the medium category and the drive category are each assigned a predetermined probability threshold, the method further comprising:

performing a response action, wherein the response action comprises:

determining that the predicted failure probability assigned to the medium category does not exceed the predetermined probability threshold assigned to the medium category and that the predicted failure probability assigned to the drive category does not exceed the predetermined probability threshold assigned to the drive category; and

allowing the queried mount event.

3. The method of claim 1 , wherein the predicted failure cause category is a category from a predefined set of failure cause categories, the predefined set of failure cause categories including a medium category associated with the queried storage medium, a drive category associated with the queried media drive, and a permit category associated with the absence of a failure cause, wherein the medium category and the drive category are each assigned a predetermined probability threshold, the method further comprising:

performing a response action, wherein the response action comprises:

determining that the predicted failure probability assigned to the medium category exceeds the predetermined probability threshold assigned to the medium category;

marking the queried storage medium as discontinued; and

sending a medium failure notification to a computing device of a monitoring center.

4. The method of claim 3 , wherein the predefined set of failure cause categories further includes a combination category associated with the combination of the queried storage medium and the queried media drive, the method further comprising:

preventing, in response to determining that a predicted failure probability assigned to the combination category exceeds a predetermined probability threshold assigned to the combination category, the queried mount event for the combination of the queried storage medium and the queried media drive.

5. The method of claim 3 , further comprising:

receiving, from the output of the machine learning model, a classification path assigned to the predicted failure probability;

storing in a path statistics file the classification path assigned to the predicted failure probability.

6. The method of claim 1 , wherein providing the machine learning model comprises:

receiving one or more training sets, each of the one or more training sets including log data of past mount events, wherein a past mounting event is an event where a storage medium is mounted by a media drive and assigned a predicted failure cause category; and

generating, by executing a learning algorithm on the one or more training sets, the machine learning model.

7. The method of claim 6 , wherein the learning algorithm includes a decision tree.

8. The method of claim 7 , the decision tree is a classification and regression tree.

9. The method of claim 6 , further comprising:

storing the input data in a database.

10. The method of claim 9 , wherein the log data of the queried storage medium and the queried media drive includes a mount quality category, the mount quality category being selected from a set of mount quality categories, the method further comprising:

generating a quality statistics file, wherein the quality statistics file includes a counter value for each mount quality category in the set of mount quality categories, the generation of the quality statistics file comprising:

reading the mount quality category for the past mount events of the queried storage medium by the queried media drive stored in the database, and

increasing the counter value of a read mount quality category; and

providing the quality statistics file with the first prediction.

11. The method of claim 9 , wherein the log data of the queried storage medium and the queried media drive includes a mount quality category, the mount quality category being selected from a set of mount quality categories, the method further comprising:

generating a quality statistics file, wherein the quality statistics file includes a medium-centered counter value for each mount quality category in the set of mount quality categories and a drive-centered counter value for each mount quality category in the set of mount quality categories, the generation of the quality statistics file comprising:

reading in the database the mount quality category for past mount events of the queried storage medium by a plurality of different media drives,

increasing the medium-centered counter value of each read mount quality category,

reading in the database the mount quality category for a past mount event of a plurality of different storage media by the queried media drive, and

increasing the drive-centered counter value of each read mount quality category; and

providing the quality statistics file with the first prediction.

12. The method of claim 9 , further comprising:

receiving one or more updated training sets, each updated training set including log data of past mount event; and

generating, by executing the learning algorithm on the one or more updated training sets, an updated machine learning model.

13. The method of claim 12 , further comprising:

generating, in response to detecting a failure event in the log data of the past mount event, a new updated training set.

14. The method of claim 1 , further comprising:

receiving a request for mounting the queried storage medium; and

selecting, from a plurality of media drives, a first media drive for mounting the queried storage medium, the selection of the first media drive comprising:

providing, for each of the plurality of media drives, the input data for a combination of the queried storage medium and the queried media drive,

receiving a second prediction for the combination, the second prediction including a mount quality value,

ranking the media drives by the mount quality value in first order and by total mount count of each media drive in second order, and

selecting the highest-ranked idle media drive as the first media drive.

15. The method of claim 14 , wherein the mount quality value includes the predicted failure probability.

16. A system for failure prediction of a queried mount event when mounting a queried storage medium by a queried media drive, the system comprising:

a memory; and

a processor in communication with the memory, the processor being configured to perform operations comprising:

providing a machine learning model having an input and an output;

receiving input data, the input data including log data of the queried storage medium and the queried media drive, wherein the input data is generated by forcing a memory dump while the queried storage medium is mounted by the queried media drive, and by a memory readout while the storage medium is stored outside of the queried media drive;

providing the input data to the input of the machine learning model;

determining, from the output of the machine learning model, a predicted failure cause category and a predicted failure probability assigned to the predicted failure cause category; and

providing a first prediction to a user.

17. The system of claim 16 , wherein the predicted failure cause category is a category from a predefined set of failure cause categories, the predefined set of failure cause categories including a medium category associated with the queried storage medium, a drive category associated with the queried media drive, and a permit category associated with the absence of a failure cause, wherein the medium category and the drive category are each assigned a predetermined probability threshold, the operations further comprising:

performing a response action, wherein the response action comprises:

determining that the predicted failure probability assigned to the medium category does not exceed the predetermined probability threshold assigned to the medium category and that the predicted failure probability assigned to the drive category does not exceed the predetermined probability threshold assigned to the drive category; and

allowing the queried mount event.

18. A computer program product for failure prediction of a queried mount event when mounting a queried storage medium by a queried media drive, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method, the method comprising:

providing a machine learning model having an input and an output;

receiving input data, the input data including log data of the queried storage medium and the queried media drive, wherein the input data is generated by forcing a memory dump while the queried storage medium is mounted by the queried media drive, and by a memory readout while the storage medium is stored outside of the queried media drive;

providing the input data to the input of the machine learning model;

determining, from the output of the machine learning model, a predicted failure cause category and a predicted failure probability assigned to the predicted failure cause category; and

providing a first prediction to a user.

19. The computer program product of claim 18 , wherein the predicted failure cause category is a category from a predefined set of failure cause categories, the predefined set of failure cause categories including a medium category associated with the queried storage medium, a drive category associated with the queried media drive, and a permit category associated with the absence of a failure cause, wherein the medium category and the drive category are each assigned a predetermined probability threshold, the method further comprising:

performing a response action, wherein the response action comprises:

determining that the predicted failure probability assigned to the medium category exceeds the predetermined probability threshold assigned to the medium category;

marking the queried storage medium as discontinued; and

sending a medium failure notification to a computing device of a monitoring center.