International Business Machines Corporation

Sweta Singh

Vaibhav Murlidhar Kulkarni

Manish Anand

Pathan Niyazulla Khan

Eitan D Farchi

Mario D Briggs

Aiswarya L Mohanasundaram

Srikkanth R Kulkarni

Mario Dominic Savio Briggs

Deepak Anil Mahajan

Harshita Pandey

Devyanshi Singh

Missula Meghana

Renu P Sumam

Dynamically determining timing for a machine learning model reward signal by receiving, by a first machine learning model, data associated with a first system state, and an action determined by a second machine learning model according to the first system state, determining, by the first machine learning model, a sleep time duration for the second machine learning model according to the first system state and the action, receiving, by the first machine learning model, after the sleep time duration, data associated with a second system state, determining, by the first machine learning model, a first reward signal according to the difference between the first system state and the second system state and the sleep time duration and updating the first machine learning model according to the first reward signal, the difference between the first system state and the second system state, and the sleep time duration.

Determining reinforcement learning model reward signal timing

Provided is a method for method for encrypting log file data in a multitenant database. The method comprises receiving a request to secure data of a tenant in a multitenant database. The method further comprises obtaining a symmetric encryption key that is specific to the tenant. The method further comprises encrypting data of the tenant using the symmetric encryption key. The data that is encrypted is stored in the multitenant database. The method further comprises encrypting a set of log file entries using the symmetric encryption key. The set of log file entries are associated with data of the tenant.

Encryption of database log files

One or more computer processors identify one or more similar, historical regression tests and historical builds utilizing a computed similarity measure between a regressed build and one or more historical builds conducted on a same release cycle, wherein the identified one or more similar historical regression tests and historical builds are K closest neighbors to the regressed build; predict an elapsed time of the one or more profiled regression tests utilizing a KNN algorithm comprising the K closest neighbors each weighted by a corresponding average distance from a test point and the elapsed time as a target variable; responsive to the predicted elapsed time exceeding an actual elapsed time associated with the regressed build, determine that the regressed build is an actual regression; responsive to determining that the regressed build is not due to variability, apply one or more mitigation actions to the regressed build.

Mitigating regression test variability

Embodiments are for generating a digital signature of a query execution plan using similarity hashing. A technique includes generating a node digital signature for nodes in a query and generating an edge digital signature for edges in the query, the edges connecting the nodes. The technique includes selecting at least one previously executed query based on the node digital signature and the edge digital signature for the query and causing the query to be processed according to an assignment associated with the at least one previously executed query.

Generate digital signature of a query execution plan using similarity hashing

Some embodiments of the present invention are directed towards techniques for validating performance degradation of cloud deployed application from neighbor based variability. Historical runs of an application deployed in a cloud environment are received. In these embodiments, a subset of these historical runs, using associated performance metrics recorded during the historical runs, are compared against performance metrics of a current version of the application which is deployed in a cloud environment to determine a subset of historical runs similar to the current version. The determined subset is then used to draw comparisons with performance metrics of a baseline run of the application to validate if a performance degradation has occurred by updating the application to the current version, reducing the impact of neighbor-based variability on evaluating performance degradation. Detailed performance analysis resources are less likely to be expended on performance degradation caused by neighbor activities in the cloud environment.

Detecting performance degradation in remotely deployed applications

A method, computer program product, and system for analyzing performance of a computer system where at a plurality of time intervals, a set of KPIs of a database and database performance outputs are monitored. A database performance value is calculated based on the monitored KPIs and one or more database performance outputs. In response to determining the database performance value is below a first threshold, a KPI correlation coefficient is calculated. The calculated KPI correlation coefficient with the greatest value is selected as a target KPI. In response to determining that a count of functions of the target KPI meets a second threshold, a dynamic tabular representation of the determined target KPI is generated. A correlation matrix is generated based on the generated tabular representation and the plurality of time intervals and a first set of target functions are identified based on the generated correlation matrix.

Root cause analysis of performance problems

A method for analyzing performance of a computer system where at a plurality of time intervals, a set of KPIs of a database and database performance outputs are monitored. A database performance value is calculated based on the monitored KPIs and one or more database performance outputs. In response to determining the database performance value is below a first threshold, a KPI correlation coefficient is calculated. The calculated KPI correlation coefficient with the greatest value is selected as a target KPI. In response to determining that a count of functions of the target KPI meets a second threshold, a dynamic tabular representation of the determined target KPI is generated. A correlation matrix is generated based on the generated tabular representation and the plurality of time intervals and a first set of target functions are identified based on the generated correlation matrix.

Embodiments of the present invention relate to a method, computer program product and system for optimizing database transactions configured for receiving a query. The query specifies a set of predicates supplied to the query and a minimal number of predicates to be satisfied for the query to be true. An operation using the query is performed on a repository that is stored in a computer readable storage medium. A set of results satisfying the minimal number of predicated is rendered.

Optimizing database query

Provided are a computer implemented method, computer program product, and system for re-writing a query. A query including a query predicate is received. It is determined that the query predicate includes one column with an associated predicate mapper. An equivalent predicate is generated for the query predicate using the predicate mapper that provides a mapping from the one column to another column. A cost of the query with the equivalent predicate is evaluated. In response to determining that the cost of the query is optimal, the query is rewritten to replace the query predicate with the equivalent predicate.

Optimizing queries using predicate mappers

A first query is received including a logical expression as a set of predicates in Disjunctive Normal Form. Each predicate from is represented as a bitwise predicate pattern to generate a set of bitwise predicate patterns. A set of valid bitwise patterns from the set of bitwise predicate patterns is identified using the first query. The set of valid bitwise patterns is parsed using logical operators to generate a logical bitwise expression. The logical bitwise expression is factored and the factored logical bitwise expression is mapped to corresponding predicates to generate a logical predicate expression. A second query is generated, which includes modifying the first query using the logical predicate expression and at least one logical operator. A subset of data is retrieved from the data repository using the second query.

Growing community of inventors

Bengaluru, India

Sweta Singh