公开(公告)号：US20240330296A1

公开(公告)日：2024-10-03

申请号：US18128152

申请日：2023-03-29

Applicant: Snowflake Inc.

Inventor： Benoit Dageville ,  Lyuping Du ,  Martin Hentschel ,  Elkhan Mammedov ,  Lukas Moll ,  Vikram Wakade

IPC: G06F16/2455 ,  G06F16/23

CPC classification number: G06F16/24552 ,  G06F16/2358 ,  G06F16/2365

Abstract: Techniques for active invalidation of cache entries are described. A first timestamp and a second timestamp are stored by a compute instance, the first timestamp corresponding to an initial time the computing instance read from a version update log in a database and the second timestamp corresponding to a last time the computing instance read from the version update log in the database. The compute instance can read records from the version update log that are associated with version timestamps greater than the second timestamp. The compute instance can invalidate a first set of cache entries with older versions stored in a cache based on reading the records, retaining a second set of cache entries, and then execute a query using the second set of cache entries.

公开(公告)号：US12210527B2

公开(公告)日：2025-01-28

申请号：US18128152

申请日：2023-03-29

Applicant: Snowflake Inc.

Inventor： Benoit Dageville ,  Lyuping Du ,  Martin Hentschel ,  Elkhan Mammedov ,  Lukas Moll ,  Vikram Wakade

IPC: G06F7/00 ,  G06F16/23 ,  G06F16/2455 ,  G06F17/00

Abstract: Techniques for active invalidation of cache entries are described. A first timestamp and a second timestamp are stored by a compute instance, the first timestamp corresponding to an initial time the computing instance read from a version update log in a database and the second timestamp corresponding to a last time the computing instance read from the version update log in the database. The compute instance can read records from the version update log that are associated with version timestamps greater than the second timestamp. The compute instance can invalidate a first set of cache entries with older versions stored in a cache based on reading the records, retaining a second set of cache entries, and then execute a query using the second set of cache entries.