RUN FORMATION IN LARGE SCALE SORTING USING BATCHED REPLACEMENT SELECTION

Patent №

US 6,571,244

Granted

2003-05-27

Filed 1999

Owner

MICROSOFT CORPORATION

AI components

1

kr

Assignment

Recorded

Dataset

AIPD

2023_r1 edition

Application

09437714

A large-scale sorting process utilizes a batched replacement selection method to form runs of sorted data records. The batched replacement selection method involves reading multiple records from a persistent data storage into main memory and sorting the multiple records to form a mini-run of multiple sorted data records. After formation, the mini-run is added to a selection tree by inserting a pointer to a first record in the mini-run into the array of pointers. The first record is linked to remaining records in the mini-run. As records are selected for output from the selection tree, the methodology replaces the selected record with a next record in the associated mini-run (if not empty) or alternatively deletes the node if the mini-run is empty. The selected records are collected into an output buffer. When the number of records reaches a pre-determined number, the selected records are written in batch back to the persistent data storage.

Knowledge representationG06F 16/9027Y10S 707/99937Y10S 707/99944

AI classification

Knowledge representation0.97
Machine learning0.47
AI hardware0.07
Planning0.00
Evolutionary computation0.00
Natural language0.00
Speech0.00
Vision0.00

Ownership

MICROSOFT CORPORATION

assignment · 103890366

Assignors

LARSON, PER-AKE

On an employer assignment, the assignors are typically the inventors.

© 2026 NYSGPT2525 LLC