System and method for selecting a backup plan based on sampled metadata

1. A method for performing backup of electronic data, the method comprising:

sampling metadata of files on a computer during a predetermined time interval, wherein the files are backed up to a storage server in accordance with a first backup plan;

extracting derivative features from attributes of the metadata by:

analyzing values in the attributes of the metadata; and

encoding the values analyzed into encoded values in a particular format for the derivative features, wherein the derivative features are created based on different types of values in the metadata;

predicting file usage dynamics for a time interval subsequent to the predetermined time interval, based on a combination of the derivative features and the attributes of the metadata, wherein the predicted file usage dynamics indicate, for the time interval, how the files are utilized on the computer, and one or more of: which of the files are subjected to particular file operations more often than other files, which of the files grow in a monotonic manner, which of the files are periodically accessed, and which of the files are un-periodically accessed;

comparing the predicted file usage dynamics for the time interval with file usage dynamics of the predetermined time interval; and

in response to determining, based on the comparison, that a difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval exceeds a threshold difference:

selecting a second backup plan for the files; and

backing up the files to the storage server in accordance with the second backup plan.

2. The method of claim 1 , further comprising:

bounding the encoded values according to the particular format; and

extracting the derivative features from the attributes in the metadata according to the encoded values.

3. The method of claim 1 , further comprising:

in response to determining, based on the comparison, that the difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval does not exceed the threshold difference, not selecting a new backup plan for the files.

4. The method of claim 1 , further comprising:

in response to determining, based on the comparison, that the difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval exceeds the threshold difference:

deriving a model of change of the metadata over time, wherein the model of change is a snapshot of dynamically sampled attributes of the files and reflects a state of the computer over the predetermined time interval.

5. The method of claim 4 , further comprising:

predicting, based on the model of change, a speed of change in size of files of a particular file type and a change in capacity of the computer subsequent to the predetermined time interval;

generating an urgency model based on the metadata further indicating (1) how soon a file must be retrieved from backup and (2) conditions on whether to back up the file in low latency storage or high latency storage;

identifying on the computer at least one newly stored file of a first file type;

selecting a third backup plan for the at least one newly stored file, the third backup plan indicating when and where to perform a backup according to a determined urgency of the at least one newly stored file, a predicted speed of change in size of files of the first file type, and the predicted change in capacity of the computer.

6. The method of claim 5 , further comprising one or more of:

predicting urgency of the at least one newly stored file on the computer based on the urgency model; and

predicting attributes of the at least one newly stored file on the computer based on the urgency model.

7. The method of claim 6 , wherein the predicted attributes comprise one or more of file type and file size.

8. A system for performing backup of electronic data, the system comprising:

a hardware processor configured to:

sample metadata of files on a computer during a predetermined time interval, wherein the files are backed up to a storage server in accordance with a first backup plan;

extract derivative features from attributes of the metadata by:

analyzing values in the attributes of the metadata; and

encoding the values analyzed into encoded values in a particular format for the derivative features, wherein the derivative features are created based on different types of values in the metadata;

predict file usage dynamics for a time interval subsequent to the predetermined time interval, based on a combination of the derivative features and the attributes of the metadata, wherein the predicted file usage dynamics indicate, for the time interval, how the files are utilized on the computer, and one or more of: which of the files are subjected to particular file operations more often than other files, which of the files grow in a monotonic manner, which of the files are periodically accessed, and which of the files are un-periodically accessed;

compare the predicted file usage dynamics for the time interval with file usage dynamics of the predetermined time interval; and

in response to determining, based on the comparison, that a difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval exceeds a threshold difference:

select a second backup plan for the files; and

back up the files to the storage server in accordance with the second backup plan.

9. The system of claim 8 , wherein the hardware processor is further configured to:

bound the encoded values according to the particular format; and

extract the derivative features from the attributes in the metadata according to the encoded values.

10. The system of claim 8 , wherein the hardware processor is further configured to:

in response to determining, based on the comparison, that the difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval does not exceed the threshold difference, not select a new backup plan for the files.

11. The system of claim 8 , wherein the hardware processor is further configured to:

in response to determining, based on the comparison, that the difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval exceeds the threshold difference:

derive a model of change of the metadata over time, wherein the model of change is a snapshot of dynamically sampled attributes of the files and reflects a state of the computer over the predetermined time interval.

12. The system of claim 11 , wherein the hardware processor is further configured to:

predict, based on the model of change, a speed of change in size of files of a particular file type and a change in capacity of the computer subsequent to the predetermined time interval;

generate an urgency model based on the metadata further indicating (1) how soon a file must be retrieved from backup and (2) conditions on whether to back up the file in low latency storage or high latency storage;

identify on the computer at least one newly stored file of a first file type;

select a third backup plan for the at least one newly stored file, the third backup plan indicating when and where to perform a backup according to a determined urgency of the at least one newly stored file, a predicted speed of change in size of files of the first file type, and the predicted change in capacity of the computer.

13. The system of claim 12 , wherein the hardware processor is further configured to perform one or more of:

predicting urgency of the at least one newly stored file on the computer based on the urgency model; and

predicting attributes of the at least one newly stored file on the computer based on the urgency model.

14. The system of claim 13 , wherein the predicted attributes comprise one or more of file type and file size.

15. A non-transitory computer-readable medium storing thereon instructions that when executed perform a method for backup of electronic data, the instructions comprising:

sampling metadata of files on a computer during a predetermined time interval, wherein the files are backed up to a storage server in accordance with a first backup plan;

extracting derivative features from attributes of the metadata by:

analyzing values in the attributes of the metadata; and

encoding the values analyzed into encoded values in a particular format for the derivative features, wherein the derivative features are created based on different types of values in the metadata;

predicting file usage dynamics for a time interval subsequent to the predetermined time interval, based on a combination of the derivative features and the attributes of the metadata, wherein the predicted file usage dynamics indicate, for the time interval, how the files are utilized on the computer, and one or more of:

which of the files are subjected to particular file operations more often than other files, which of the files grow in a monotonic manner, which of the files are periodically accessed, and which of the files are un-periodically accessed;

comparing the predicted file usage dynamics for the time interval with file usage dynamics of the predetermined time interval; and

in response to determining, based on the comparison, that a difference between the predicted file usage dynamics and the file usage dynamics of the predetermined time interval exceeds a threshold difference:

selecting a second backup plan for the files; and

backing up the files to the storage server in accordance with the second backup plan.