Virtual warehouse analysis and configuration planning system

Patent metadata

Field	Value
Publication number	US-12468999-B2
Application number	US-202318213670-A
Country	US
Kind code	B2
Filing date	Jun 23, 2023
Priority date	Jun 23, 2023
Publication date	Nov 11, 2025
Grant date	Nov 11, 2025

Abstract

Methods, systems, and apparatuses for using machine learning to simulate changes to virtual warehouse configurations without access to data stored by corresponding virtual warehouses are described herein. A computing device may receive first performance metrics of one or more first queries executed by one or more first virtual warehouses. The computing device may then generate a trained machine learning model to simulate operating parameter changes and predict virtual warehouse query performance metrics. The computing device may then provide performance metrics for one or more second virtual warehouses to the trained machine learning model. Output from the trained machine learning model may comprise performance metric predictions corresponding to a given configuration of a virtual warehouse. Predicted costs associated with those performance metric predictions may be output and, based on user input, the operating parameter of the at least one of the one or more second virtual warehouses may be modified.

First claim

What is claimed is:

1 . A computing device configured to use machine learning to simulate changes to virtual warehouse configurations without access to data stored by corresponding virtual warehouses, the computing device comprising:
one or more processors; and memory storing instructions that, when executed by the one or more processors, cause the computing device to:
collect first performance metrics of one or more first queries during execution, by one or more first virtual warehouses, of the one or more first queries, wherein each of the one or more first virtual warehouses comprises a respective set of computing resources configured to:
execute one or more queries with respect to at least a portion of a plurality of data warehouses,
collect results from the one or more queries, and
provide access to the collected results;

generate a trained machine learning model by training, using the first performance metrics, an artificial neural network comprising a plurality of nodes to simulate operating parameter changes and predict virtual warehouse query performance metrics, wherein training the artificial neural network comprises modifying, based on the first performance metrics, one or more weights of the plurality of nodes;
collect second performance metrics of one or more second queries during execution, by one or more second virtual warehouses different from the one or more first virtual warehouses, of the one or more second queries;
provide, as input to the trained machine learning model, the second performance metrics;
receive, as output from the trained machine learning model and in response to the input to the trained machine learning model, data indicating:
first performance metric predictions corresponding to a first configuration for an operating parameter of at least one of the one or more second virtual warehouses; and
second performance metric predictions corresponding to a second configuration for the operating parameter of the at least one of the one or more second virtual warehouses;

cause display, in an interface, of:
a first predicted cost based on the first performance metric predictions and associated with the first configuration, and
a second predicted cost based on the second performance metric predictions and associated with the second configuration;

receive, via the interface, a selection of an option corresponding to the first configuration;
modify, based on the selection, the operating parameter of the at least one of the one or more second virtual warehouses by transmitting, to a virtual warehouse management application, instructions that prevent the one or more second virtual warehouses from executing one or more types of queries during a particular period of time defined by a schedule;
receive, via a second interface and after modification of the operating parameter, input indicating feedback corresponding to the at least one of the one or more second virtual warehouses; and
further train, based on the input and based on the selection, the trained machine learning model by modifying, based on the feedback corresponding to the at least one of the one or more second virtual warehouses, at least one weight of the one or more weights of the plurality of nodes.

2 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to modify the operating parameter of the at least one of the one or more second virtual warehouses by causing the computing device to:
modify one or more of:
a size of the one or more second virtual warehouses;
a schedule of the one or more second virtual warehouses;
a minimum number of clusters of the one or more second virtual warehouses;
a maximum number of clusters of the one or more second virtual warehouses;
an auto suspend time of the one or more second virtual warehouses;
a statement timeout of the one or more second virtual warehouses;
a query acceleration setting of the one or more second virtual warehouses; or
a setting that controls whether the one or more second virtual warehouses are optimized for an application programming interface (API).

3 . The computing device of claim 1 , wherein the one or more types of queries correspond to queries received from one or more users.

4 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to generate the trained machine learning model by causing the computing device to:
train the machine learning model based on third performance metrics of one or more third queries executed by the one or more second virtual warehouses.

5 . The computing device of claim 1 , wherein the feedback corresponds to a cost of the at least one of the one or more second virtual warehouses.

6 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, cause the computing device to:
instantiate, based on the selection, an additional virtual warehouse.

7 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, further cause the computing device to:
receive, after the computing device modifies the operating parameter, an indication of a requested query; and output a recommended virtual warehouse of the one or more second virtual warehouses for executing the requested query.

8 . The computing device of claim 1 , wherein the instructions, when executed by the one or more processors, further cause the computing device to provide, as the input to the trained machine learning model, the second performance metrics by causing the computing device to:
receive, via the user interface, a selection of the first configuration, wherein the second performance metrics comprises an indication of the first configuration.

9 . A method for using machine learning to simulate changes to virtual warehouse configurations without access to data stored by corresponding virtual warehouses, the method comprising:
collecting, by a computing device, first performance metrics of one or more first queries during execution, by one or more first virtual warehouses, of the one or more first queries, wherein each of the one or more first virtual warehouses comprises a respective set of computing resources configured to:
execute one or more queries with respect to at least a portion of a plurality of data warehouses,
collect results from the one or more queries, and
provide access to the collected results;
generating, by the computing device, a trained machine learning model by training, using the first performance metrics, an artificial neural network comprising a plurality of nodes to simulate operating parameter changes and predict virtual warehouse query performance metrics, wherein training the artificial neural network comprises modifying, based on the first performance metrics, one or more weights of the plurality of nodes;
collecting second performance metrics of one or more second queries during execution, by one or more second virtual warehouses different from the one or more first virtual warehouses, of the one or more second queries;
providing, by the computing device and as input to the trained machine learning model, the second performance metrics;
receiving, by the computing device and as output from the trained machine learning model in response to the input to the trained machine learning model, data indicating:
first performance metric predictions corresponding to a first configuration for an operating parameter of at least one of the one or more second virtual warehouses; and
second performance metric predictions corresponding to a second configuration for the operatin

Assignees

• Capital One Services Llc

Inventors

• Syal Rahul
• Fearns Tyler
• Sylvia Chris
• Bougher Justin
• Kolanu Kishore
• Ellis David
• Mallen Brenton
• Khopkar Rohan

Classifications

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06Q10/0639 Primary

  Performance analysis of employees; Performance analysis of enterprise or organisation operations · CPC title

• G06F16/283 Primary

  Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP · CPC title