# US 2021/0334282 A1 · Prescriptive Analytics Based NOSQL Database Service Optimization System for Cloud Computing

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

- Number: US 2021/0334282 A1
- Document ID: US20210334282A1
- Status: Published application
- Kind: A1
- Application number: 16/896,681
- Filed: 2020-06-09
- Published: 2021-10-28
- Earliest claimed: 2020-04-28
- Assignees: Accenture Global Solutions Limited
- Applicants: Accenture Global Solutions Limited
- Inventors: Madhan Kumar Srinivasan; Guruprasad PV; Samba Sivachari Rage
- Web page: https://patentgrep.com/patent/US20210334282A1

## CPC

- G06F 16/24542: {Plan optimisation}
- G06F 16/903: Querying (for retrieval from the web G06F16/953)
- G06F 16/217: {Database tuning (G06F16/2282 takes precedence; database performance monitoring G06F11/3409)}
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/09: Supervised learning
- 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 system includes a multi-layer throughput optimization (TPO) stack to generate a token containing prescriptions for rightsizing database service throughput. A prescriptive engine layer of the TPO stack may obtain consumed throughputs from the utilization tracking data; predict, based on the consumed throughputs, a prediction-based throughput for the target NoSQL database service; select target consumed throughputs from the consumed throughputs based on a time granularity; determine a statistical throughput consumption value from the target consumed throughputs; calculate a statistics-based throughput over a TPO interval based on the statistical throughput consumption value; extract database service performance metric data from the utilization tracking data; select target performance metric data from the database service performance metric data based on the time granularity; determine a statistical performance metric value from the target performance metric data; select one of the prediction-based throughput and the statistics-based throughput as based on the statistical performance metric value.

## Claims

### Claim 1

A system comprising:
network interface circuitry configured to:
receive utilization tracking data for a target NoSQL database service provisioned with provisioned throughputs, a provisioned throughput representing a capacity of the target NoSQL database service to process data per unit time; and
send a throughput optimization (TPO) token to a host interface, the host interface configured to execute the TPO token to provision the target NoSQL database service with a recommended throughput;
optimization circuitry in data communication with the network interface circuitry, the optimization circuitry configured to execute a TPO stack, the TPO stack executable to:
at a prescriptive engine layer of the TPO stack:
obtain consumed throughputs for each unit time from the utilization tracking data;
predict, based on the consumed throughputs, a prediction-based throughput over a TPO interval for the target NoSQL database service with a prediction engine;
select target consumed throughputs from the consumed throughputs based on a predetermined time granularity;
determine a statistical throughput consumption value from the target consumed throughputs, wherein the statistical throughput consumption value is greater than consumed throughput values of a predetermined portion of the target consumed throughputs;
calculate a statistics-based throughput over the TPO interval for the target NoSQL database service based on the statistical throughput consumption value;
extract database service performance metric data for each unit time from the utilization tracking data;
select target performance metric data from the database service performance metric data based on the predetermined time granularity;
determine a statistical performance metric value from the target performance metric data, wherein the statistical performance metric value is greater than performance metric values of a predetermined portion of the target performance metric data;
select one of the prediction-based throughput and the statistics-based throughput as the recommended throughput based on the statistical performance metric value, a performance metric threshold, and a prediction accuracy of the prediction engine;
perform a throughput determination to execute a throughput optimization for the target NoSQL database service based on the recommended throughput; and
generate the TPO token based on the throughput determination.

### Claim 2 (depends on 1)

The system of claim 1, wherein the optimization circuitry is configured to:
determine the statistical throughput consumption value as a Nth percentile consumed throughput value of the target consumed throughputs; and
determine the statistical performance metric value as a Nth percentile performance metric value of the target performance metric data;
wherein the Nth percentile is greater than or equal to a 95th percentile.

### Claim 3 (depends on 1)

The system of claim 1, wherein the target performance metric data comprises rate limited request records, error rate records, and throughput failed request records, and the optimization circuitry is configured to:
calculate a Nth percentile rate limited request value from the rate limited request records, a Nth percentile error rate value from the error rate records, and a Nth percentile throughput failed request value from the throughput failed request records; and
select one of the prediction-based throughput and the statistics-based throughput as the recommended throughput based on at least one of the Nth percentile rate limited request value, the Nth percentile error rate value, the Nth percentile throughput failed request value, and a prediction accuracy of the prediction engine.

### Claim 4 (depends on 3)

The system of claim 3, wherein the optimization circuitry is configured to select the prediction-based throughput as the recommended throughput in response to the Nth percentile rate limited request value being less than or equal to a rate limited request threshold, the Nth percentile error rate value being less than or equal to an error rate threshold, the Nth percentile throughput failed request value being less than or equal to a throughput failed request threshold, and a prediction accuracy being greater than an accuracy threshold.

### Claim 5 (depends on 4)

The system of claim 4, wherein the optimization circuitry is configured to select the statistics-based throughput as the recommended throughput in response to the Nth percentile rate limited request value being less than or equal to the rate limited request threshold, the Nth percentile error rate value being less than or equal to the error rate threshold, the Nth percentile throughput failed request value being less than or equal to the throughput failed request threshold, and the prediction accuracy being less than or equal to the accuracy threshold.

### Claim 6 (depends on 5)

The system of claim 5, wherein the optimization circuitry is configured to:
in response to neither the prediction-based throughput nor the statistics-based throughput being selected as the recommended throughput,
calculate an average provisioned throughput from the provisioned throughputs,
add a provisioned buffer to the average provisioned throughput to obtain a buffered provisioned throughput, and
select the buffered provisioned throughput as the recommended throughput.

### Claim 7 (depends on 6)

The system of claim 6, wherein the optimization circuitry is further configured to generate a consumption optimization preview for the target NoSQL database service based on the recommended throughput, the average provisioned throughput, and consumption metrics data for the target NoSQL database service.

### Claim 8 (depends on 7)

The system of claim 7, wherein the optimization circuitry is further configured to determine a priority to execute the throughput optimization for the target NoSQL database service based on the consumption optimization preview.

### Claim 9 (depends on 1)

The system of claim 1, wherein the optimization circuitry is further configured to:
select target provisioned throughputs from the provisioned throughputs based on the predetermined time granularity, each of the target provisioned throughputs having a maximum provisioned throughput value among provisioned throughputs within respective time intervals with the predetermined time granularity;
calculate a Nth percentile provisioned throughput value from the target provisioned throughputs;
calculate a throughput consumption ratio based on the Nth percentile provisioned throughput value and the statistical throughput consumption value;
determine a prediction-based throughput buffer to add to the prediction-based throughput based on the throughput consumption ratio;
add the prediction-based throughput buffer to the prediction-based throughput to obtain a first buffered throughput;
determine a statistics-based throughput buffer to add to the statistics-based throughput based on the throughput consumption ratio;
add the statistics-based throughput buffer to the statistics-based throughput to obtain a second buffered throughput; and
select one of the first buffered throughput and the second buffered throughput as the recommended throughput based on the statistical performance metric value and a prediction accuracy of the prediction engine.

### Claim 10 (depends on 1)

The system of claim 1, wherein:
the TPO stack is further executable to, at a presentation layer of the TPO stack, generate a TPO control interface including a TPO-window presentation; and
the TPO-window presentation includes a selectable option to execute the throughput optimization.

### Claim 11 (depends on 10)

The system of claim 10, wherein the TPO-window presentation further includes a selectable option to reject execution of the throughput optimization.

### Claim 12 (depends on 10)

The system of claim 10, wherein the TPO-window presentation is configured to display multiple throughput optimizations, each of the respective throughput optimizations for a corresponding database service.

### Claim 13 (depends on 1)

The system of claim 1, wherein the optimization circuitry is further configured to:
calculate a database service volume usage ratio based on a provisioned database service volume on the target NoSQL database service and a consumed database service storage volume by the target NoSQL database service;
determine a recommended database service storage volume for the target NoSQL database service based on the database service volume usage ratio;
perform a storage volume determination to execute a storage volume optimization for the target NoSQL database service based on the recommended database service storage volume; and
generate the TPO token based on the throughput determination and the storage volume determination.

### Claim 14 (depends on 1)

The system of claim 1, wherein the prediction engine includes a long short-term memory neural network model trained to predict throughput of the target NoSQL database service.

### Claim 15 (depends on 1)

The system of claim 1, wherein the target NoSQL database service is a NoSQL database or a NoSQL container in the NoSQL database.

### Claim 16 (depends on 1)

The system of claim 1, wherein the optimization circuitry is further configured to:
select the target NoSQL database service from a plurality of candidate database services by:
obtaining a utilization time length for a candidate database service;
obtaining a utilization data missing coefficient for the candidate database service; and
determination of the candidate database service as the target NoSQL database service by comparison of the utilization time length to a utilization time length threshold and comparison of the utilization data missing coefficient to a data missing tolerance threshold.

### Claim 17 (depends on 1)

The system of claim 1, wherein each of the target consumed throughputs has a maximum consumed throughput value among consumed throughputs within respective time intervals with the predetermined time granularity and each of the target performance metric data has a maximum performance metric value among database service performance metric data within respective time intervals with the predetermined time granularity.

### Claim 18 (depends on 17)

The system of claim 17, wherein the TPO stack is further executable to:
at an input layer of the TPO stack:
obtain, via the network interface circuitry, the utilization tracking data; and
store the utilization tracking data via a database service operation at a data-staging layer of the TPO stack;
at a configuration layer of the TPO stack:
determine the TPO interval; and
determine the performance metric threshold.

### Claim 19

A method comprising:
receiving, at network interface circuitry, utilization tracking data for a target NoSQL database service provisioned with provisioned throughputs, a provisioned throughput representing a capacity of the target NoSQL database service to process data per unit time;
executing a throughput optimization (TPO) stack with optimization circuitry, the optimization circuitry in data communication with the network interface circuitry, executing the TPO stack comprising:
obtaining, at a prescriptive engine layer of the TPO stack, consumed throughputs for each unit time from the utilization tracking data;
predict, at the prescriptive engine layer, based on the consumed throughputs, a prediction-based throughput over a TPO interval for the target NoSQL database service with a prediction engine;
select, at the prescriptive engine layer, target consumed throughputs from the consumed throughputs based on a predetermined time granularity;
determine, at the prescriptive engine layer, a statistical throughput consumption value from the target consumed throughputs, wherein the statistical throughput consumption value is greater than consumed throughput values of a predetermined portion of the target consumed throughputs;
calculate, at the prescriptive engine layer, a statistics-based throughput over the TPO interval for the target NoSQL database service based on the statistical throughput consumption value;
extract, at the prescriptive engine layer, database service performance metric data for each unit time from the utilization tracking data;
select, at the prescriptive engine layer, target performance metric data from the database service performance metric data based on the predetermined time granularity;
determine, at the prescriptive engine layer, a statistical performance metric value from the target performance metric data, wherein the statistical performance metric value is greater than performance metric values of a predetermined portion of the target performance metric data;
select, at the prescriptive engine layer, one of the prediction-based throughput and the statistics-based throughput as a recommended throughput based on the statistical performance metric value, a performance metric threshold, and a prediction accuracy of the prediction engine;
perform, at the prescriptive engine layer, a throughput determination to execute a throughput optimization for the target NoSQL database service based on the recommended throughput; and
generate, at the prescriptive engine layer, a TPO token based on the throughput determination; and
sending, via the network interface circuitry, the TPO token to a host interface configured to execute the TPO token to provision the target NoSQL database service with the recommended throughput.

### Claim 20

A product including:
machine-readable media other than a transitory signal; and instructions stored on the machine-readable media, the instructions configured to, when executed, cause a machine to:
at network interface circuitry:
receive utilization tracking data for a target NoSQL database service provisioned with provisioned throughputs, a provisioned throughput representing a capacity of the target NoSQL database service to process data per unit time;
at optimization circuitry in data communication with the network interface circuitry, the optimization circuitry executing a throughput optimization (TPO) stack:
obtain, at a prescriptive engine layer of the TPO stack, consumed throughputs for each unit time from the utilization tracking data;
predict, at the prescriptive engine layer, based on the consumed throughputs, a prediction-based throughput over a TPO interval for the target NoSQL database service with a prediction engine;
select, at the prescriptive engine layer, target consumed throughputs from the consumed throughputs based on a predetermined time granularity;
determine, at the prescriptive engine layer, a statistical throughput consumption value from the target consumed throughputs, wherein the statistical throughput consumption value is greater than consumed throughput values of a predetermined portion of the target consumed throughputs;
calculate, at the prescriptive engine layer, a statistics-based throughput over the TPO interval for the target NoSQL database service based on the statistical throughput consumption value;
extract, at the prescriptive engine layer, database service performance metric data for each unit time from the utilization tracking data;
select, at the prescriptive engine layer, target performance metric data from the database service performance metric data based on the predetermined time granularity;
determine, at the prescriptive engine layer, a statistical performance metric value from the target performance metric data, wherein the statistical performance metric value is greater than performance metric values of a predetermined portion of the target performance metric data;
select, at the prescriptive engine layer, one of the prediction-based throughput and the statistics-based throughput as a recommended throughput based on the statistical performance metric value, a performance metric threshold, and a prediction accuracy of the prediction engine;
perform, at the prescriptive engine layer, a throughput determination to execute a throughput optimization for the target NoSQL database service based on the recommended throughput; and
generate, at the prescriptive engine layer, a TPO token based on the throughput determination; and
send, via the network interface circuitry, the TPO token to a host interface configured to execute the TPO token to provision the target NoSQL database service with the recommended throughput.