Systems and methods for artificial intelligence-based data system optimization
OwnerAngelo Katroulisinventor
Life of this patent
Estimating the term…
Abstract
In some aspects, the disclosure is directed to methods and systems for access path selection optimization for relational database queries. Machine learning systems, such as neural networks, may be used to determine costs for each path. A neural network may be trained based on cost determinations from scans and indexes of the tree for an initial set of queries, and then may be used to predict costs for additional queries such that a path may be selected. The training data may be periodically refreshed or updated, or may be refreshed or updated responsive to changes in hardware or computing environment.
Claims20
A system for machine learning-based optimization of a data system, comprising:
- a computing system comprising one or more memory devices storing a knowledge store comprising tuples of a set of features and observed characteristics of a query response, and one or more processors executing:
- a learned optimizer configured to estimate a cost for each of a plurality of alternate query methodologies for performing a given query of a database maintained by one or more computing device, and select a lowest cost query methodology;
- an executor in communication with the database, the executor configured to perform the given query according to the selected lowest cost query methodology;
- and
- a learning core configured to generate a tuple of the knowledge store from a response to the given query, and comprising a neural network configured to predict the characteristics for each query response;
- and
- wherein the learned optimizer utilizes model weights of the neural network to estimate the costs for each of the plurality of alternate query methodologies.
A method for machine learning-based optimization of a data system, comprising:
- estimating, by a learned optimizer of a computing system, a cost for each of a plurality of alternate query methodologies for performing a given query of a database from one or more tuples comprising a set of features and observed characteristics of query responses, the one or more tuples generated by a learning core of the computing system and stored in a knowledge store comprising one or more memory devices, the database maintained by one or more computing devices;
- selecting, by the learned optimizer, a lowest cost query methodology based on the estimated costs;
- and
- executing, by an executor of the computing system in communication with the database, the given query according to the selected query methodology.
A non-transitory computer readable medium comprising instructions that, when executed by one or more processors of a computing system, cause the computing system to:
- estimate a cost for each of a plurality of alternate query methodologies for performing a given query of a database from one or more tuples comprising a set of features and observed characteristics of query responses, the one or more tuples generated by a learning core of the computing system and stored in a knowledge store comprising one or more memory devices, the database maintained by one or more computing devices;
- select a lowest cost query methodology based on the estimated costs;
- and
- execute the given query according to the selected query methodology.
A method for machine learning-based optimization of a data system, comprising:
- for each of a plurality of candidate worker threads, estimating, by a learned optimizer of a computing system, a cost to execute a modification of a stored data state, the cost estimated according to a neural network model trained from historical observations of costs of data state modifications by the candidate worker threads;
- selecting, by the learned optimizer, a lowest cost candidate worker thread to perform the modification of the stored data state based on the estimated costs;
- and
- executing, by an executor of the computing system using the selected worker thread, the modification.
Dependencies are inferred from the claim text.