# US 2023/0273908 A1 · Automated mainframe database maintenance

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2023/0273908 A1
- Document ID: US20230273908A1
- Status: Published application
- Kind: A1
- Application number: 17/680,558
- Filed: 2022-02-25
- Published: 2023-08-31
- Earliest claimed: 2022-02-25
- Assignees: Kyndryl, Inc.
- Applicants: Kyndryl, Inc.
- Inventors: Marcelo De Souza Vaz; Marcos Eugenio Fernandes; Thiago Bianchi; Adriana Melges Quintanilha Weingart
- Web page: https://patentgrep.com/patent/US20230273908A1

## CPC

- G06F 16/217: {Database tuning (G06F16/2282 takes precedence; database performance monitoring G06F11/3409)}
- G06F 11/004: {Error avoidance (G06F11/07 and subgroups take precedence)}
- G06F 11/0727: {in a storage system, e.g. in a DASD or network based storage system (drivers for digital recording or reproducing units G06F3/06; circuits for error detection or correction within digital recording or reproducing units G11B20/18; for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS], H04L67/1097)}
- G06F 11/3409: {for performance assessment}
- G06F 16/2272: {Management thereof}
- G06F 16/2282: {Tablespace storage structures; Management thereof}
- G06F 16/2308: {Concurrency control (transaction processing G06F9/466)}
- G06F 16/2477: {Temporal data queries}
- G06N 3/08: Learning methods
- G06F 2201/80: Database-specific techniques
- G06N 20/00: Machine learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Systems and methods for automated mainframe database maintenance are provided. In implementations, a method includes obtaining, by a computing device, real-time performance metrics of a mainframe database; automatically generating, by the computing device, a predicted maintenance task as an output of a trained database maintenance task classification machine learning (ML) model based on an input of the real-time performance metrics; automatically generating, by the computing device, a time to execute the predicted maintenance task as an output of a trained database maintenance triggering ML model based on an input of the predicted maintenance task and the real-time performance metrics; automatically generating, by the computing device, maintenance task instructions for the mainframe database based on the predicted maintenance task, the time to execute the predicted maintenance task, and a maintenance profile of the mainframe database; and automatically initiating, by the computing device, the execution of the maintenance task instructions.

## Claims

### Claim 1

A method, comprising:
obtaining, by a computing device, real-time performance metrics of a mainframe database;
automatically generating, by the computing device, a predicted maintenance task as an output of a trained database maintenance task classification machine learning (ML) model based on an input of the real-time performance metrics;
automatically generating, by the computing device, a time to execute the predicted maintenance task as an output of a trained database maintenance triggering ML model based on an input of the predicted maintenance task and the real-time performance metrics;
automatically generating, by the computing device, maintenance task instructions for the mainframe database based on the predicted maintenance task, the time to execute the predicted maintenance task, and a maintenance profile of the mainframe database; and
automatically initiating, by the computing device, the execution of the maintenance task instructions by the mainframe database.

### Claim 2 (depends on 1)

The method of claim 1, wherein:
the database maintenance task classification ML model is trained by correlating historic performance related metrics of a time series dataset with a historic maintenance data profile of the time series dataset utilizing a first classification machine learning technique; and
the database maintenance triggering ML model is trained with the historic performance related metrics of the time series dataset and the historic maintenance data profile of the time series dataset utilizing a first regression machine learning technique.

### Claim 3 (depends on 2)

The method of claim 2, wherein, the historic performance related metrics include the following three sets of time-series historical variables: variables related to a mainframe performance, variables related to an organizational database aspect, and variables related to access plans.

### Claim 4 (depends on 3)

The method of claim 3, wherein:
the variables related to a mainframe performance include: average transaction response time, central processing unit (CPU) usage, and number of transactions considered in a predefined time period;
the variables related to an organizational database aspect include: table space availability, index space availability, and table fragmentation rate; and
the variables related to access plans include: table size, table type, number of indexes, types of indexes, and number of columns.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
automatically generating, by the computing device, an indication of an anomalous behavior as an output of a trained database monitor anomaly detection ML model based on an input of the real-time performance metrics, wherein the database monitor anomaly detection ML model is trained with the historic performance related metrics of the time series dataset utilizing a second regression machine learning technique; and
automatically sending, by the computing device, an alert indicating the anomalous behavior to a remote client device via a network connection.

### Claim 6 (depends on 1)

The method of claim 1, further comprising;
automatically generating, by the computing device, a probability value of an imminent failure as an output of a trained database maintenance failure profile ML model based on an input of the real-time performance metrics, wherein the database maintenance failure profile ML model is trained by correlating the historic performance related metrics of the time series dataset with a historic failure data profile of the time series dataset utilizing a second classification ML technique;
automatically determining, by the computing device, an imminent failure in response to determining that the probability value is greater than a predetermined failure threshold value;
automatically generating, by the computing device, a suggested recovery profile as an output of a trained database maintenance failure profile ML model based on an input of the imminent failure, wherein the database maintenance failure profile ML model is trained by correlating the historic performance related metrics of the time series dataset with a historic recovery profile of the time series data; and
automatically sending, by the computing device, the suggested recovery profile to a remote client device of a user via a network connection.

### Claim 7 (depends on 1)

The method of claim 1, further comprising performing the training of the database maintenance task classification ML model and the database maintenance triggering ML model.

### Claim 8 (depends on 1)

The method of claim 1, further comprising automatically generating, by the computing device, the time series dataset based on catalog data from a mainframe catalog of the mainframe database.

### Claim 9 (depends on 1)

The method of claim 1, wherein the maintenance task instructions are executed without a requirement to restart the mainframe database.

### Claim 10

A computer program product comprising one or more computer readable storage media having program instructions collectively stored on the one or more computer readable storage media, the program instructions executable to:
obtain real-time performance metrics of a mainframe database;
automatically generate a predicted maintenance task as an output of a trained database maintenance task classification machine learning (ML) model based on an input of the real-time performance metrics;
automatically generate a time to execute the predicted maintenance task as an output of a trained database maintenance triggering ML model based on an input of the predicted maintenance task and the real-time performance metrics;
automatically generate maintenance task instructions for the mainframe database based on the predicted maintenance task, the time to execute the predicted maintenance task, and a maintenance profile of the mainframe database; and
automatically initiate the execution of the maintenance task instructions by the mainframe database.

### Claim 11 (depends on 10)

The computer program product of claim 10, wherein:
the database maintenance task classification ML model is trained by correlating historic performance related metrics of a time series dataset with a historic maintenance data profile of the time series dataset utilizing a first classification machine learning technique; and
the database maintenance triggering ML model is trained with the historic performance related metrics of the time series dataset and the historic maintenance data profile of the time series dataset utilizing a first regression machine learning technique.

### Claim 12 (depends on 11)

The computer program product of claim 11, wherein the historic performance related metrics include the following three sets of time-series historical variables: variables related to a mainframe performance, variables related to an organizational database aspect, and variables related to access plans.

### Claim 13 (depends on 12)

The computer program product of claim 12, wherein:
the variables related to a mainframe performance include are selected from one or more of the group consisting of: average transaction response time, central processing unit (CPU) usage, and number of transactions considered in a predefined time period;
the variables related to an organizational database aspect are selected from one or more of the group consisting of: table space availability, index space availability, and table fragmentation rate; and
the variables related to access plans are selected from one or more of the group consisting of: table size, table type, number of indexes, types of indexes, and number of columns.

### Claim 14 (depends on 10)

The computer program product of claim 10, wherein the program instructions are further executable to:
automatically generate an indication of an anomalous behavior as an output of a trained database monitor anomaly detection ML model based on an input of the real-time performance metrics, wherein the database monitor anomaly detection ML model is trained with the historic performance related metrics of the time series dataset utilizing a second regression machine learning technique; and
automatically send an alert indicating the anomalous behavior to a remote client device via a network connection.

### Claim 15 (depends on 10)

The computer program product of claim 10, wherein the program instructions are further executable to:
automatically generate a probability value of an imminent failure as an output of a trained database maintenance failure profile ML model based on an input of the real-time performance metrics, wherein the database maintenance failure profile ML model is trained by correlating the historic performance related metrics of the time series dataset with a historic failure data profile of the time series dataset utilizing a second classification ML technique;
automatically determine an imminent failure in response to determining that the probability value is greater than a predetermined failure threshold value;
automatically generate a suggested recovery profile as an output of a trained database maintenance failure profile ML model based on an input of the imminent failure, wherein the database maintenance failure profile ML model is trained by correlating the historic performance related metrics of the time series dataset with a historic recovery profile of the time series data; and
automatically send the suggested recovery profile to a remote client device of a user via a network connection.

### Claim 16 (depends on 10)

The computer program product of claim 10, wherein the program instructions are further executable to perform the training of the database maintenance task classification ML model and the database maintenance triggering ML model.

### Claim 17 (depends on 10)

The computer program product of claim 10, wherein the program instructions are further executable to automatically generate the time series dataset based on catalog data from a mainframe catalog of the mainframe database.

### Claim 18

A system comprising:
a processor, a computer readable memory, one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instructions executable to:
obtain real-time performance metrics of a mainframe database;
automatically generate a predicted maintenance task as an output of the trained database maintenance task classification ML model based on an input of the real-time performance metrics;
automatically generate a time to execute the predicted maintenance task as an output of a trained database maintenance triggering ML model based on an input of the predicted maintenance task and the real-time performance metrics;
automatically generate maintenance task instructions for the mainframe database based on the predicted maintenance task, the time to execute the predicted maintenance task, and a maintenance profile of the mainframe database; and
automatically initiate the execution of the maintenance task instructions by the mainframe database.

### Claim 19 (depends on 18)

The system of claim 18, wherein:
automatically generate a time series dataset based on catalog data from a mainframe catalog of a mainframe database;
train the database maintenance task classification machine learning (ML) model by correlating historic performance related metrics of the time series dataset with a historic maintenance data profile of the time series dataset utilizing a first classification machine learning technique; and
train the database maintenance triggering ML model with the historic performance related metrics of the time series dataset and the historic maintenance data profile of the time series dataset utilizing a first regression machine learning technique,
wherein the historic performance related metrics include the following three sets of time-series historical variables: variables related to a mainframe performance, variables related to an organizational database aspect, and variables related to access plans.

### Claim 20 (depends on 19)

The system of claim 19, wherein the program instructions are further executable to:
automatically generate an indication of an anomalous behavior as an output of a trained database monitor anomaly detection ML model based on an input of the real-time performance metrics, wherein the database monitor anomaly detection ML model is trained with the historic performance related metrics of the time series dataset utilizing a second regression machine learning technique;
automatically send an alert indicating the anomalous behavior to a remote client device via a network connection in response to the generation of the indication of the anomalous behavior;
automatically generate a probability value of an imminent failure as an output of a trained database maintenance failure profile ML model based on an input of the real-time performance metrics, wherein the database maintenance failure profile ML model is trained by correlating the historic performance related metrics of the time series dataset with a historic failure data profile of the time series dataset utilizing a second classification ML technique;
automatically determine an imminent failure in response to determining that the probability value is greater than a predetermined failure threshold value;
automatically generate a suggested recovery profile as an output of a trained database maintenance failure profile ML model based on the determining the imminent failure, wherein the database maintenance failure profile ML model is trained by correlating the historic performance related metrics of the time series dataset with a historic recovery profile of the time series data; and
automatically send the suggested recovery profile to a remote client device of a user via a network connection.
