Database redo log optimization by skipping MVCC redo log records

    公开(公告)号:US10915413B2

    公开(公告)日:2021-02-09

    申请号:US15410701

    申请日:2017-01-19

    Applicant: SAP SE

    Abstract: Restoring of a database table in the database system (e.g., an in-memory insert-only database system, etc.) is initiated. Thereafter, a transaction log volume storing data log records and multi-version concurrency control (MVCC) log records corresponding to the database table is accessed. Based on such access, data log records corresponding to the database table are replayed while insert and update MVCC log records corresponding to the database table are skipped. Subsequently, restoring of the database table in the database system is finalized. Related apparatus, systems, techniques and articles are also described.

    SPARSE INFRASTRUCTURE FOR TRACKING AD-HOC OPERATION TIMESTAMPS

    公开(公告)号:US20200301906A1

    公开(公告)日:2020-09-24

    申请号:US16893567

    申请日:2020-06-05

    Applicant: SAP SE

    Abstract: Disclosed herein are system, method, and computer program product embodiments for managing timestamp information in memory systems. In an embodiment, an infrastructure may utilize a controller to manage sparse timestamp blocks. These sparse timestamp blocks may manage timestamp information in a persistent memory structure. Controller 110 may utilize a transient timestamp accessor to map the timestamp information for faster hash lookups. Controller 110 may also utilize a garbage collection map as a bitmap to further save processing steps. Controller 110 may utilize the sparse timestamp blocks, transient timestamp accessor, and garbage collection map to efficiently store timestamp data and to quickly retrieve the stored timestamp data.

    Database processing after a lock condition

    公开(公告)号:US10706019B2

    公开(公告)日:2020-07-07

    申请号:US14861438

    申请日:2015-09-22

    Applicant: SAP SE

    Abstract: Disclosed herein are system, method, and computer program product embodiments for performing database processing after a lock condition. An embodiment operates by receiving a release of a lock on data of a database. It is determined whether the data was modified by a processing of a first transaction associated with the first request. The lock is provided to a second transaction associated with a second request to access the data. If the first transaction did not modify the data, a second transaction associated with the second request is retried without refreshing a snapshot of the data. If the first transaction modified the data, the snapshot of the data is refreshed before the second transaction is retried.

    Sparse infrastructure for tracking ad-hoc operation timestamps

    公开(公告)号:US10678772B2

    公开(公告)日:2020-06-09

    申请号:US15980974

    申请日:2018-05-16

    Applicant: SAP SE

    Abstract: Disclosed herein are system, method, and computer program product embodiments for managing timestamp information in memory systems. In an embodiment, an infrastructure may utilize a controller to manage sparse timestamp blocks. These sparse timestamp blocks may manage timestamp information in a persistent memory structure. Controller 110 may utilize a transient timestamp accessor to map the timestamp information for faster hash lookups. Controller 110 may also utilize a garbage collection map as a bitmap to further save processing steps. Controller 110 may utilize the sparse timestamp blocks, transient timestamp accessor, and garbage collection map to efficiently store timestamp data and to quickly retrieve the stored timestamp data.

    Time slider operator for temporal data aggregation

    公开(公告)号:US10515059B2

    公开(公告)日:2019-12-24

    申请号:US15665154

    申请日:2017-07-31

    Applicant: SAP SE

    Abstract: Calculation of aggregated values in a history database table can be optimized using an approach in which an ordered history table is accessed. The ordered history table can include a sequential listing of commit identifiers associated with updates, insertions, and/or deletions to values in the database table. The ordered history table can be traversed in a single pass to calculate an aggregation function using an optimized algorithm. The optimized algorithm can enable calculation of an aggregated metric of the values based on a selected method for tracking invalidated values to their corresponding commit identifiers. The calculated metric is generated for a current version of the database table; and promoted.

    Database Redo Log Optimization by Skipping MVCC Redo Log Records

    公开(公告)号:US20180203771A1

    公开(公告)日:2018-07-19

    申请号:US15410701

    申请日:2017-01-19

    Applicant: SAP SE

    CPC classification number: G06F11/1471 G06F16/211 G06F2201/80

    Abstract: Restoring of a database table in the database system (e.g., an in-memory insert-only database system, etc.) is initiated. Thereafter, a transaction log volume storing data log records and multi-version concurrency control (MVCC) log records corresponding to the database table is accessed. Based on such access, data log records corresponding to the database table are replayed while insert and update MVCC log records corresponding to the database table are skipped. Subsequently, restoring of the database table in the database system is finalized. Related apparatus, systems, techniques and articles are also described.

Patent Agency Ranking