AI-based online recovery of non-critical business process data with referential integrity
A machine learning (ML) module that intelligently stores data in an online backup system based on the criticality of a business process. For a critical business process, all pre- and post-operation data are stored in the backup. For a non-critical business process, the ML module does not collect all data, but collects that data which deviates from a pre-defined target data pattern. Such deviating data is stored for a given time period and then discarded if no recovery request is received for it within the time period. Subsequent receipt of similar type of data may trigger the ML module to not store it. Thus, online backup storage is maintained at a manageable size without any loss of data for a critical business process. During data recovery, data is retrieved from the online backup with referential integrity to restore the original database without downtime to give a truly online recovery experience.
1. A method comprising:
receiving, by a computing system, a first indication that a first business process is a non-critical business process;
determining, by the computing system, that a first database entry for a first operation associated with the first business process deviates from a first pre-defined target pattern;
storing, by the computing system, the first database entry for a pre-determined time period, wherein the first database entry is initially stored in a database, and wherein the method further comprises:
receiving, by the computing system, historical data from the database associated with the first business process, wherein the historical data includes content related to a plurality of transactions associated with the first business process prior to the first operation;
training, by the computing system, a machine learning (ML) module with the content of the historical data to enable the ML module to recognize a set of pre-defined data patterns present in the content and relationships among the data patterns, wherein the set of pre-defined data patterns includes the pre-defined target pattern; and
generating, by the computing system, a trained version of the ML module based on the training thereof; and
discarding, by the computing system, the stored first database entry in the absence of a first recovery request for the first operation within the pre-determined time period.
2. The method of claim 1 , wherein storing the first database entry comprises:
storing, by the computing system, the first database entry in a backup storage different from a database in which the first database entry was originally stored.
3. The method of claim 1 , further comprising:
in the absence of the first recovery request, recognizing, by the computing system, a data pattern of the first database entry as equivalent to the first pre-defined target pattern; and
beyond the pre-determined time period, discontinuing, by the computing system, storage of those database entries associated with the first business process that contain data patterns similar to the data pattern of the first database entry.
4. The method of claim 1 , wherein the first database entry includes one or more of the following:
date and time of the first operation;
an identification of a transaction associated with the first operation;
an identification of an originator of the transaction; and
data associated with the transaction.
5. The method of claim 1 , wherein the determining comprises:
using, by the computing system, the trained version of the ML module to determine that the first database entry for the first operation deviates from the pre-defined target pattern.
6. The method of claim 1 , wherein a second database entry for a second operation associated with the first business process is initially stored in a database, and wherein the method further comprises:
determining, by the computing system, that the second database entry deviates from a second pre-defined target pattern;
storing, by the computing system, the second database entry in a backup storage that is different from the database;
receiving, by the computing system, a second recovery request for the second operation within the pre-determined time period; and
performing, by the computing system, data recovery related to the second recovery request from the backup storage to the database based on data stored in the backup storage prior to the second operation.
7. The method of claim 6 , further comprising:
identifying, by the computing system, a data pattern of the second database entry; and
continuing, by the computing system, to store in the backup storage and beyond the pre-determined time period those database entries associated with the first business process that contain data patterns similar to the data pattern of the second database entry.
8. The method of claim 6 , wherein the first database entry and the second database entry are stored in a JavaScript Object Notation (JSON) format.
9. The method of claim 1 , further comprising:
receiving, by the computing system, a second indication that a second business process is a critical business process; and
storing, by the computing system, a second database entry for a second operation associated with the second business process beyond the pre-determined time period regardless of whether the second database entry deviates from a second pre-defined target pattern.
10. A computing system comprising:
a memory storing program instructions; and
a processing unit coupled to the memory and operable to execute the program instructions, which, when executed by the processing unit, cause the computing system to:
receive a first indication that a first business process is a non-critical business process;
determine that a first database entry for a first operation associated with the first business process deviates from a first pre-defined target pattern;
store the first database entry for a pre-determined time period, wherein the first database entry is initially stored in a database, and wherein the program instructions, upon execution by the processing unit, cause the computing system to:
receive historical data from the database associated with the first business process, wherein the historical data includes content related to a plurality of transactions associated with the first business process prior to the first operation;
train a machine learning (ML) module with the content of the historical data to enable the ML module to recognize a set of pre-defined data patterns present in the content and relationships among the data patterns, wherein the set of pre-defined data patterns includes the pre-defined target pattern;
generate a trained version of the ML module; and
use the trained version of the ML module to determine that the first database entry for the first operation deviates from the pre-defined target pattern; and
discard the stored first database entry in the absence of a first recovery request for the first operation within the pre-determined time period.
11. The computing system of claim 10 , wherein the program instructions, upon execution by the processing unit, cause the computing system to store the first database entry in a backup storage different from a database in which the first database entry was originally stored.
12. The computing system of claim 10 , wherein a second database entry for a second operation associated with the first business process is initially stored in a database, and wherein the program instructions, upon execution by the processing unit, cause the computing system to:
determine that the second database entry deviates from a second pre-defined target pattern;
store the second database entry in a backup storage that is different from the database;
receive a second recovery request for the second operation within the pre-determined time period;
perform data recovery related to the second recovery request from the backup storage to the database based on data stored in the backup storage prior to the second operation;
identify a data pattern of the second database entry; and
continue to store in the backup storage and beyond the pre-determined time period those database entries associated with the first business process that contain data patterns similar to the data pattern of the second database entry.
13. The computing system of claim 10 , wherein the program instructions, upon execution by the processing unit, cause the computing system to:
receive a second indication that a second business process is a critical business process; and
store a second database entry for a second operation associated with the second business process beyond the pre-determined time period regardless of whether the second database entry deviates from a second pre-defined target pattern.
14. The computing system of claim 13 , wherein the program instructions, upon execution by the processing unit, cause the computing system to store the first database entry and the second database entry in a JavaScript Object Notation (JSON) format.
15. A computer program product comprising a non-transitory computer-usable medium having computer-readable program code embodied therein, the computer-readable program code adapted to be executed by a computing system to implement a method comprising:
receiving an indication that a first business process is a non-critical business process;
determining that a first database entry for a first operation associated with the first business process deviates from a first pre-defined target pattern;
storing the first database entry for a pre-determined time period;
receiving historical data from a database associated with the first business process, wherein the first database entry is initially stored in the database, and wherein the historical data includes content related to a plurality of transactions associated with the first business process prior to the first operation;
training a machine learning (ML) module with the content of the historical data to enable the ML module to recognize a set of pre-defined data patterns present in the content and relationships among the data patterns, wherein the set of pre-defined data patterns includes the pre-defined target pattern;
generating a trained version of the ML module based on the training thereof; and
using the trained version of the ML module to determine that the first database entry for the first operation deviates from the pre-defined target pattern; and
discarding the stored first database entry in the absence of a first recovery request for the first operation within the pre-determined time period.
16. The computer program product of claim 15 , wherein the method further comprises:
in the absence of the first recovery request, recognizing a data pattern of the first database entry as equivalent to the first pre-defined target pattern; and
beyond the pre-determined time period, discontinuing storage of those database entries associated with the first business process that contain data patterns similar to the data pattern of the first database entry.
17. The computer program product of claim 15 , wherein a second database entry for a second operation associated with the first business process is initially stored in a database, and wherein the method further comprises:
determining that the second database entry deviates from a second pre-defined target pattern;
storing the second database entry in a backup storage that is different from the database;
receiving a second recovery request for the second operation within the pre-determined time period; and
performing data recovery related to the second recovery request from the backup storage to the database based on data stored in the backup storage prior to the second operation.