Snowflake Inc.

Shrainik Jain

Jiaqi Yan

Allison Waingold Lee

Stratis Viglas

Qiuye Jin

Yongsik Yoon

Boyung Lee

Prasanna Rajaperumal

Nicola Dan Onose

Prashant Gaharwar

Karan Chadha

The subject technology receives a query, the query including a statement for performing the query. The subject technology performs a lookup operation on a stored plan cache based on the query. The subject technology performs, in response to a cache match of the query to a stored query plan in the stored plan cache based on the lookup operation, a validation process of the stored query plan. The subject technology determines whether the stored query plan is valid based on the validation process. The subject technology performs, in response to determining that the stored query plan is valid, a program building process for the stored query plan to generate a final query plan. The subject technology sends the final query plan to an execution node for execution.

Optimized validation of cached query plans

A system to identify optimal cloud resources for executing workloads. The system deduplicates historical client queries based on a workload selection configuration to determine a grouping of historical client queries. The system generates a workload based on at least a portion of the grouping of historical client queries. The system repeatedly executes a test run of the workload using resources of a cloud environment to determine whether there is a performance difference in the test run. The system, in response to determining that there is no performance difference, identifies one or more sets of decreased resources of the cloud environment. The system re-executes the test run using the one or more sets of decreased resources of the cloud environment to determine whether there is a performance difference in the test run that is attributed to the one or more sets of decreased resources of the cloud environment.

Optimized identification of performance regression

Queries executed against a materialized view can execute up to orders of magnitude faster than equivalent queries on a source (or base) table. However, although a query can reference a materialized view directly, a user (e.g., query author) may not know about a relevant materialized view. Moreover, if a source table has multiple materialized views generated, the user may not know which materialized view to reference in the query. Thus, embodiments of the present disclosure provide techniques for automatically rewriting queries directed to a source table to utilize existing materialized views.

Transient materialized view rewrite

A system to repeatedly execute a test run of a workload using resources of a cloud environment to determine whether there is a performance difference in the test run. The system to, in response to determining that there is no performance difference, identify one or more sets of decreased resources of the cloud environment. The system to re-execute the test run using the one or more sets of decreased resources of the cloud environment to determine whether there is a performance difference in the test run that is attributed to the one or more sets of decreased resources of the cloud environment. The system to determine minimum resources of the cloud environment to repeatedly execute the test run using the minimum resources without existence of a performance difference in response to re-executing the test run using the one or more sets of decreased resources of the cloud environment.

Identification of optimal cloud resources for executing workloads

Systems, methods, and devices for feature release and workload capture in database systems are disclosed. The method includes determining a workload based on one or more client queries to be rerun to test a feature that is unreleased to one or more database clients. The method includes repeatedly executing a test run of the workload to determine a stability factor of the test run. The method includes re-executing, in response to determining the stability factor of the test run, the test run using resources with a different concurrency to confirm the stability factor of the test run. The method includes releasing the feature to the one or more database clients in response to confirming the stability factor of the test run.

Feature release and workload capture in database systems

Systems, methods, and devices for incremental feature development and workload capture in database systems are disclosed. The method includes determining a workload based on one or more client queries to be rerun for testing the feature release. The method further includes repeatedly executing a test run of the workload to identify whether there is a performance difference in the test run. The method further includes re-execute, in response to identifying the performance difference, the test run using resources with a different concurrency to identify whether the performance difference still exists. The method further includes flagging the performance difference as a false positive in response to identifying that the performance difference no longer exists when the test run is re-executed using the resources with different concurrency.

Incremental feature development and workload capture in database systems

A baseline run of a workload and a target run of the workload are compared to identify whether there is a performance regression in the target run, wherein the target run implements a feature that is not implemented in the baseline run. In response to identifying the performance regression, the target run is executed to identify whether the performance regression still exists. The performance regression is flagged as a false positive in response to identifying that the performance regression no longer exists when the target run is executed.

Systems, methods, and devices for incremental feature development and workload capture in database systems are disclosed. A method includes determining a workload having one or more historical client queries to be rerun for testing the feature, wherein the feature comprises procedural logic. The method further includes executing a baseline run of the workload that does not implement the feature and executing a target run of the workload while implementing the feature. The method further includes comparing the baseline run and the target run to identify whether there is a performance regression in the target run. The method further includes, in response to identifying the performance regression, rerunning the target run to identify whether the performance regression still exists.

Growing community of inventors

Seattle, WA, United States of America

Shrainik Jain