# US 2023/0055415 A1 · Machine learning and optimization techniques for job scheduling

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

- Number: US 2023/0055415 A1
- Document ID: US20230055415A1
- Status: Published application
- Kind: A1
- Application number: 17/408,078
- Filed: 2021-08-20
- Published: 2023-02-23
- Earliest claimed: 2021-08-20
- Assignees: Kyndryl, Inc.
- Applicants: Kyndryl, Inc.
- Inventors: Robert Bradley DESAULNIERS; Clea Anne ZOLOTOW; Mihai CRIVETI; Ana Maria BEZERRA MAIMONI
- Web page: https://patentgrep.com/patent/US20230055415A1

## CPC

- G06F 11/3433: {for load management (allocation of a server based on load conditions G06F9/505; load rebalancing G06F9/5083; redistributing the load in a network by a load balancer H04L67/1029)}
- G06N 3/092: Reinforcement learning
- G06F 11/3006: {where the computing system is distributed, e.g. networked systems, clusters, multiprocessor systems (multiprogramming arrangements G06F9/46; allocation of resources G06F9/50)}
- G06F 11/3037: {where the computing system component is a memory, e.g. virtual memory, cache (accessing, addressing or allocating within memory systems or architectures G06F12/00; checking stores for correct operation G11C29/00)}
- G06F 11/3409: {for performance assessment}
- G06F 11/3442: {for planning or managing the needed capacity}
- G06F 18/2113: by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/08: Learning methods
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method includes training a recurrent neural network by monitoring data in a memory of a first server as the first server executes jobs and by determining an amount of computing resources used by the first server while executing the jobs and applying the recurrent neural network to data in the memory to predict an amount of computing resources that the first server will use when executing a first future job. The method also includes, in response to determining that execution of the first future job did not meet a performance criterion, making a change to the first server. The method further includes further training the recurrent neural network using a reinforcement learning technique, applying the recurrent neural network to determine that the change should be made to a second server, and in response, making the change to the second server before the second server executes a second future job.

## Claims

### Claim 1

A method comprising:
training a recurrent neural network by monitoring data in a memory of a first server as the first server executes jobs and by determining an amount of computing resources used by the first server while executing the jobs;
applying the recurrent neural network to starting data in the memory to predict an amount of computing resources that the first server will use when executing a first future job;
assigning the first future job to execute on the first server based on the predicted amount of computing resources;
in response to determining that execution of the first future job did not meet a performance criterion, making a change to the first server of the computing system;
in response to determining that execution of the first future job after making the change to the first server did meet the performance criterion, further training the recurrent neural network using a reinforcement learning technique and based on the change;
applying the recurrent neural network to determine that the change should be made to a second server before the second server executes a second future job and in response, making the change to the second server before the second server executes the second future job; and
assigning the second future job to execute on the second server.

### Claim 2 (depends on 1)

The method of claim 1, wherein training the recurrent neural network using the reinforcement learning technique comprises assigning a score to the change in response to determining that execution of the first future job after making the change did meet the performance criterion.

### Claim 3 (depends on 2)

The method of claim 2, wherein determining that the change should be made to the second server is based on the score.

### Claim 4 (depends on 1)

The method of claim 1, wherein assigning the first future job to the first server is in response to determining that the first server has the predicted amount of computing resources available.

### Claim 5 (depends on 1)

The method of claim 1, wherein the performance criterion comprises at least one of execution completion, a carbon footprint, or a cloud utilization.

### Claim 6 (depends on 1)

The method of claim 1, further comprising removing a scheduled job from a job schedule in response to determining that a result of executing the scheduled job is unused.

### Claim 7 (depends on 1)

The method of claim 1, further comprising assigning a third future job to execute on the first and second servers such that the first server executes a first portion of the third future job and the second server executes a second portion of the third future job.

### Claim 8 (depends on 1)

The method of claim 1, wherein the change comprises an adjustment to the performance criterion.

### Claim 9 (depends on 1)

The method of claim 1, wherein the change comprises an upgrade to the second server.

### Claim 10

An apparatus comprising:
a memory; and
a hardware processor communicatively coupled to the memory, the hardware processor configured to:
train a recurrent neural network by monitoring data in a memory of a first server as the first server executes jobs and by determining an amount of computing resources used by the first server while executing the jobs;
apply the recurrent neural network to starting data in the memory of the first server to predict an amount of computing resources that the first server will use when executing a first future job;
assign the first future job to execute on the first server based on the predicted amount of computing resources;
in response to determining that execution of the first future job did not meet a performance criterion, make a change to the first server of the computing system;
in response to determining that execution of the first future job after make the change to the first server did meet the performance criterion, further train the recurrent neural network using a reinforcement learning technique and based on the change;
apply the recurrent neural network to determine that the change should be made to a second server of the computing system before the second server executes a second future job and in response, make the change to the second server before the second server executes the second future job; and
assign the second future job to execute on the second server.

### Claim 11 (depends on 10)

The apparatus of claim 10, wherein training the recurrent neural network using the reinforcement learning technique comprises assigning a score to the change in response to determining that execution of the first future job after making the change did meet the performance criterion.

### Claim 12 (depends on 11)

The apparatus of claim 11, wherein determining that the change should be made to the second server is based on the score.

### Claim 13 (depends on 10)

The apparatus of claim 10, wherein assigning the first future job to the first server is in response to determining that the first server has the predicted amount of computing resources available.

### Claim 14 (depends on 10)

The apparatus of claim 10, wherein the performance criterion comprises at least one of execution completion, a carbon footprint, or a cloud utilization.

### Claim 15 (depends on 10)

The apparatus of claim 10, the hardware processor further configured to remove a scheduled job from a job schedule in response to determining that a result of executing the scheduled job is unused.

### Claim 16 (depends on 10)

The apparatus of claim 10, the hardware processor further configured to assign a third future job to execute on the first and second servers of the computing system such that the first server executes a first portion of the third future job and the second server executes a second portion of the third future job.

### Claim 17 (depends on 10)

The apparatus of claim 10 wherein the change comprises an adjustment to the performance criterion.

### Claim 18 (depends on 10)

The apparatus of claim 10, wherein the change comprises an upgrade to the second server.

### Claim 19

A computer program product for job scheduling, the computer program product comprising a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to:
train a recurrent neural network by monitoring data in a memory of a first server as the first server executes jobs and by determining an amount of computing resources used by the first server while executing the jobs;
apply the recurrent neural network to starting data in the memory to predict an amount of computing resources that the first server will use when executing a first future job;
assign the first future job to execute on the first server based on the predicted amount of computing resources;
in response to determining that execution of the first future job did not meet a performance criterion, make a change to the first server of the computing system;
in response to determining that execution of the first future job after make the change to the first server did meet the performance criterion, further train the recurrent neural network using a reinforcement learning technique and based on the change;
apply the recurrent neural network to determine that the change should be made to a second server of the computing system before the second server executes a second future job and in response, make the change to the second server before the second server executes the second future job; and
assign the second future job to execute on the second server.

### Claim 20 (depends on 19)

The computer program product of claim 19, wherein training the recurrent neural network using the reinforcement learning technique comprises assigning a score to the change in response to determining that execution of the first future job after making the change did meet the performance criterion.
