Ocient Holdings LLC

Andrew Park

Ellis Mihalko Saupe

Jason Arnold

George Kondiles

Finley Jordan Lau

Sarah Kate Schieferstein

Samuel Peter Bove

Susmita Saha

Andrew Michael Bass

Alyssa Catherine Wagenmaker

Pieter Charles Jas Svenson

Pavel Yusim

Kevin Garner

A node of a database system is operable to execute a shuffle operator based on, during a first temporal period, reserving a first amount of memory, allocating, to each of a plurality of other nodes, a corresponding number of fixed-memory units based on the first amount of memory, and process first data received from the plurality of other nodes in accordance with the corresponding number of fixed-memory units allocated to the each of the plurality of other nodes. During a second temporal period after the first temporal period, the first amount of memory is updated to a second amount of memory based on comparing an available amount of memory with a configured memory threshold, an updated corresponding number of fixed-memory units is re-allocated, and second data received from the plurality of other nodes in accordance with the corresponding number of fixed-memory units is processed.

Execution of a shuffle operator via a database system based on allocating memory units

A database system is operable to determine a query that includes a join expression for execution. A set of input rows are dispersed for processing via a plurality of parallelized join processes in conjunction with executing the join expression based on identifying a first proper subset of the set of input rows based on a null-handling strategy. First rows in the first proper subset are dispersed for processing across the plurality of parallelized join processes in accordance with the null-handling strategy. second rows in a set difference between the set of input rows and the first proper subset are dispersed across the plurality of parallelized join processes in accordance with a join key-based assignment strategy. The set of input rows are processed via the plurality of parallelized join processes.

Handling null values in processing join operations during query execution

A database system is operable to execute a row dispersal operator, in conjunction with execution of a query operator execution flow for a corresponding query, to disperse a plurality of input rows across the set of parallelized operations based on determining a plurality of input data blocks that collectively includes the plurality of input rows, where each input data block in the plurality of input data blocks a corresponding proper subset of the plurality of input rows. Each data block of the plurality of input data blocks is assigned to a corresponding one of the set of parallelized operations for processing in accordance with a uniform assignment scheme, and each data block is forwarded to the corresponding one of the set of parallelized operations by reference.

Row dispersal across parallelized processes during query execution based on forwarding data blocks

A database system is operable to execute a query based on, during a first temporal period, executing a multiplexer operation in conjunction with a first mode of operation to generate a first plurality of multiplexer output, and executing a load operation upon the first plurality of multiplexer output to generate initial load operation output from the first plurality of multiplexer output. In response to a determination to transition to execution of the multiplexer operation in conjunction with a second mode of operation, the multiplexer operation is further executed in conjunction with the second mode of operation during a second temporal period to generate a second plurality of multiplexer output for processing via the load operation. The load operation is further executed upon the second plurality of multiplexer output to generate further load operation output.

Switching modes of operation of a row dispersal operation during query execution

A database system is operable to execute a query transaction to generate addendum part data updating a segment A target storage node receives a plurality of sets of buffered rows and a plurality of flush requests. The target storage nodes each of a plurality of addendum parts in response to receiving a corresponding one of a plurality of flush requests. The target node assigns a plurality of version numbers to the plurality of addendum parts. A coordinator operator execution module commits only a final addendum part having a most recent version number as the addendum part data for the segment.

Distributed generation of addendum part data for a segment stored via a database system

A database system is operable to generate an initial query operator execution flow for a corresponding query for execution that includes an extend operation serially before a column-based filtering operation indicating filtering by new column values of a new column generated via the extend operation as a function of an existing column. The initial query operator execution flow is converted into an updated query operator execution based on pushing the column-based filtering operation serially before the extend operation, and updating the column-based filtering operation to filter by existing column values of the existing column via applying an inverse function of the function. The corresponding query is executed by applying the updated query operator execution flow based on executing the column-based filtering operation upon a plurality of rows to generate a filtered subset of the plurality of rows, and executing the extend operation upon only rows in the filtered subset.

Growing community of inventors

St. Charles, IL, United States of America

Andrew Park