Teradata Us, Inc.

Ncr Corporation

Teradota Us, Inc.

Other

Teradat Us, Inc.

Douglas Paul Brown

Anita Richards

John Mark Morris

Louis Martin Burger

Thomas Patrick Julien

Bhashyam Ramesh

Frank Roderic Vandervort

Jeetendra Chaware

Todd A Walter

Stephen Andrew Brobst

Paul Laurence Sinclair

Jeffrey S Shelton

Donald Raymond Pederson

Hoa Thu Tran

Scott E Gnau

Debra Galeazzi

Manjula Koppuravuri

Caroline Ballinger

Matthew Edward Antoun

Kenneth Ray Shortes

Awny Kayed Al-Omari

P Keith Muller

William P Ward

Mohammed Yassin Al-Kateb

Sanjay Nair

Jaiprakash G Chimanchode

Arthur Vargas Lopes

Mark Sirek

Mohamed Ahmed Yassin Eltabakh

Bruce Wayne Britton

Michael G Coutts

Richard Glick

Daniel David Hoffman

Allen N Diaz

Naveen Thaliyil Sankaran

Sergio Ferragut

Choung Kim

Preeti Javaji

Prama Agarwal

Thomas Greene

Joshua I Mereminsky

James Peter Colby, Jr

Delynn Anne Kelly

Diana Lynn Thomas

Sunanda K Reddy

Lovlean Arora

Sourabh Maity

Emiran Curtmola

Kristi O'Grady

Modie Christon Smith

Ruth Gladys Fenwick

Ryan Speed

Michael Sean McIntire

Walton Dutton

Vijayasaradhi Indurthi

Christopher James Antoun

Arthur Clark

Artur Borycki

Chrisopher James Antoun

David Doyle Gilbreath

A method, apparatus and computer program product for estimating resource consumption for steps in a query execution plan for a query performed by a relational database management system (RDBMS) in a computer system. Past execution data for the steps are used to train a machine learning (ML) model and its model parameters to predict execution times for the steps. A prediction module comprised of the ML model configured by the model parameters predicts an execution time for a current step of the query execution plan for the query, based on current step information and current system load. A boosting module boosts the current step either up or down for processing by the RDBMS to meet a service level goal (SLG) for the query, based on the predicted execution time for the current step, as well as an elapsed query time, a query SLG time, and/or a query CPU time.

Estimator of resource consumption by query execution plan steps

A query is received. It is determined that the query does not fit a profile for a run-the-business set of queries, where the profile for the run-the-business set of queries excludes queries that are not routine parts of running a business and that do not require priority processing. The query is executed with a dynamically-created compute capacity that is not part of a compute capacity used to run the run-the-business set of queries.

Dynamically instantiated complex query processing

A computer system executes a database management system (DBMS). The DBMS manages a database comprised of DBMS resources. The DBMS receives a request to be executed. The request is a DBMS action to be executed using the DBMS resources. The request includes a predicate specifying a maximum cost for executing the request, and a deadline, specifying a deadline by which the request is to be completed in its execution. The DBMS determines a plurality of workloads under which the request is qualified to execute. Each workload of the plurality of workloads includes a respective set of requests that have common characteristics. Each workload of the plurality of workloads includes a respective cost criterion and a respective elapsed time criterion. The DBMS selects a selected workload from among the plurality of workloads. The selected workload has a selected cost criterion and a selected elapsed time criterion. The DBMS begins execution of the request using the selected workload.

Managing cloud pricing and what-if analysis to meet service level goals

In some examples, a system receives an input graph representation of one or more query plans for one or more database queries, and generates, by an embedding machine learning model based on the input graph representation, a feature vector that provides a distributed representation of the one or more query plans. The system determines, using the feature vector, one or more user behaviors and/or workload characteristics of one or more workloads in one or more database systems.

Query graph embedding

A data store system may include a storage device configured to store a plurality of data store tables and may include a processor in communication with the storage device. The processor may receive a plurality of requests. For each request, the processor may: (1) determine an associated workload type for the request; (2) determine a first respective rate at which the request is to be released for scheduling of execution; and (3) release the request for scheduling of execution based on the first respective rate. For each released request, the processor may: (1) determine a second respective rate based on the associated workload type at which each released request is scheduled to be executed; and (2) in response to execution being scheduled for a released request, execute the released request. A method and computer-readable medium are also disclosed.

Workload request rate control

A database system receives a query. The database system retrieves an old query execution plan (QEP), OldPlan, for the query. The database system submits the query to an optimizer. The optimizer returns a new QEP, NewPlan, for the query. The database system submits the OldPlan and the NewPlan to a machine learning classifier (ML classifier). The ML classifier predicts that executing the NewPlan will result in a performance regression as compared to executing the OldPlan. The database system executes the OldPlan instead of the NewPlan.

Reducing query optimizer plan regressions with machine learning classification

A data engine request is received on a local data system. The data engine request includes a portion of the request that is to be processed on an external data engine system. The portion is forwarded to the external data engine system and statistics for accessing external objects of the external data engine system is acquired. The statistics are evaluated for compliance with a Service Level Goal (SLG) associated with the request. Rules-based processing permits optimization and planning of the request on the local data engine system to be modified in view of the statistics received from the external data engine system to comply with the SLG. In an embodiment, actual resource utilization metrics noted during execution of the portion on the external data engine system is provided as feedback to the local data engine system for re-planning and re-optimizing the request with a modified execution plan.

Service level goal processing for multi-environment data engine requests

An apparatus, method and computer program product for estimating as-a-Service (aaS) query prices in a relational database management system (RDBMS). An optimizer of the RDBMS inserts an EXPLAIN modifier into a query, wherein the EXPLAIN modifier results in the optimizer generating a summary of a query execution plan for the query that includes one or more cost estimates for the RDBMS to perform the query. A price estimate for the query is then generated based on the cost estimates, wherein the price estimate is generated using one or more configurable pricing formulae. The price estimate is merged into the summary of the query execution plan for the query. Moreover, a price guarantee may be generated for the price estimate, wherein the price guarantee is honored when the query is subsequently invoked for execution by the RDBMS.

Estimating as-a-service query prices within optimizer explained plans

In some examples, a system receives function descriptors for different types of functions to be used when processing database queries, each function descriptor of the function descriptors comprising information relating to a respective function of the different types of functions. The system computes, based on a first function descriptor for a first function of the different types of functions, an estimate of a runtime metric associated with execution of the first function for processing a database query.

Runtime metric estimations for functions

A determination is made that a database system is resource bound resulting in a resource bound condition. Signals for the resources being bound in the database system are identified. Events associated with the signals are extracted. Events are correlated temporally to identify a time interval for which an arrival rate meter (ARM) is helpful. Database system segments are selected that effect key performance indicators associated with the identified time interval. Parameters for the selected database system segments to be deferred by the database system are estimated. The estimated parameters are incorporated into an arrival rate meter (ARM). The ARM is put into effect.

Automatically defining arrival rate meters

An apparatus, method and computer program product for physical database design and tuning in relational database management systems. A relational database management system executes in a computer system, wherein the relational database management system manages a relational database comprised of one or more tables storing data. A Deep Reinforcement Learning based feedback loop process also executes in the computer system for recommending one or more tuning actions for the physical database design and tuning of the relational database management system, wherein the Deep Reinforcement Learning based feedback loop process uses a neural network framework to select the tuning actions based on one or more query workloads performed by the relational database management system.

Physical database design and tuning with deep reinforcement learning

A machine-learning driven Database Management System (DBMS) is provided. One or more machine-learning algorithms are trained on the database constructs and execution plans produced by a database optimizer for queries. The trained machine-learning algorithms provide predictors when supplied the constructs and plans for a given query. The predictors are processed by the DBMS to make resource, scheduling, and Service Level Agreement (SLA) compliance decisions with respect to the given query.

Machine-learning driven database management

A data store system may include at least one storage device to store a plurality of data and at least one processor with access to the storage device. The at least one processor may receive a plurality of features associated with an environment. The at least one processor may further generate a state representation of the environment based on the plurality of features. The at least one processor may further generate a plurality of predicted future states of the environment based on the state representation. The at least one processor may further generate at least one action to be performed by the environment based on the plurality of predicted future states. The at least one processor may provide the at least one action to the environment to be performed. A method and computer-readable medium are also disclosed.

Autonomous workload management in an analytic platform

In a database system, at least one metric associated with resources in a database system used by multiple classes of requests is monitored, where a first of the multiple classes is associated with a lower priority than a second of the multiple classes. A throttle limit is calculated for requests of the first class, based on the monitored metric. The calculated throttle limit is used to determine scheduling of the request of the first class for execution.

Calculating a throttle limit for requests in a database system

A database system receives a query to be processed. The database system has resources. A user assigns the query to a tier of resource allocation priorities in a hierarchy of tiers. The tier has been designated as being automatically managed by the database system. The tier has a plurality of levels of priority for resource allocation (LPRAs). The database system decomposes the query into a first step and a set of subsequent steps. The first step has a beginning and each of the set of subsequent steps has a respective beginning. The database system assigns the first step to a first LPRA, wherein executing the query at the first LPRA is projected by the database system to satisfy a service level goal (SLG) within a on_schedule_range of the SLG. The database system determines during execution of the set of subsequent steps that the query is no longer projected to satisfy the SLG within the on_schedule_range of the SLG and, as a result, assigns one of the set of subsequent steps to a second LPRA different from the first LPRA, wherein executing the query at the second LPRA is projected by the database system to return execution of the query to within the on_schedule_range of the SLG.

Automatic resource allocation design for satisfying service level goals of mixed workload queries in a database system

Growing community of inventors

Rancho Santa Fe, CA, United States of America

Douglas Paul Brown