US 7,890,455 B2Granted

System and apparatus to ensure a low-latency read of log records from a database management system (“DBMS”)

AssigneeInternational Business Machines Corporation

InventorsSerge Bourbonnais, Daniel B. Morales, Clarence M. Pruet, Jonathan W. Wierenga, Tyrone C. Yu

Life of this patent

Estimating the term…

Loading patent…

Abstract

A system and method to ensure a low-latency read of log records from a Database Management System (“DBMS”) in asynchronous log-based database replication capture from a blocking log read Application Programming Interface (“API”). The system may include a replication server with a log read module to initialize a log read thread. The reading module of the log read thread may read a predefined number of DBMS log records. A log flush module of the log flush thread may be executed by the log read thread and waits for a read complete notification. The update module, in response to a log flush thread time out, may also generate a plurality of loggable transactions such that a number of log records in an internal API log satisfies a DBMS threshold and unblocks the DBMS API. Therefore, the latency time may be constant, low, and predictable ensuring an accurate replication capture.

Claims4

Dependent claims

A non-transitory computer readable medium for ensuring a low-latency read of log records from a Database Management System (“DBMS”) in asynchronous log-based database replication capture from a blocking log read Application Programming Interface (“API”), the non-transitory computer readable medium comprising:

a log read module configured to initialize a log read thread in response to a log read begin, the log read thread configured to read a predefined number of DBMS log records in a single API call, the log read thread executing a log flush thread;
a log flush module of the log flush thread configured to be executed by the log read thread, the log flush thread configured to wait for a read complete notification, the log flush thread configured to wait for a predefined time period, the predefined time period exceeding an expected log read call time;
an update module of the log flush thread configured to generate a loggable transaction, the loggable transaction updating a single row of a database table, the update module generating a plurality of loggable transactions such that a number of log records in an internal API log satisfies a DBMS threshold in response to a log flush thread time out:
wherein the update module discontinues generating a plurality of loggable transactions in response to reaching a predetermined limit, the log read thread notifying the log flush thread upon a successful read, the log flush thread completing;
a reading module of the log read thread configured to read a predefined number of DBMS log records in response to the internal API log having the number of log records satisfying the DBMS threshold, such that the log read thread has an unblocked state;
and
a notification module of the log read thread configured to notify the log flush thread of a read complete such that the log flush thread completes.

4 more elements

A system for ensuring a low-latency read of log records from a Database Management System (“DBMS”) in asynchronous log-based database replication capture from a blocking log read Application Programming Interface (“API”), the system comprising:

a network configured to communicate data between a plurality of devices;
a DBMS server in communication with the network;
a source database;
a target server in communication with the network;
a target database;
a replication server in communication with the network, the replication server comprising:
a processor;
a storage device comprising one or more data files;
an input device;
an output device;
a memory comprising:
a log read module configured to initialize a log read thread in response to a log read begin, the log read thread configured to read a predefined number of DBMS log records in a single API call, the log read thread executing a log flush thread;
a log flush module of the log flush thread configured to be executed by the log read thread, the log flush thread configured to wait for a read complete notification, the log flush thread configured to wait for a predefined time period, the predefined time period exceeding an expected log read call time;
an update module of the log flush thread configured to generate a loggable transaction, the loggable transaction updating a single row of a database table, the update module generating a plurality of loggable transactions such that a number of log records in an internal API log satisfies a DBMS threshold in response to a log flush thread time out:
wherein the update module discontinues generating a plurality of loggable transactions in response to reaching a predetermined limit, the log read thread notifying the log flush thread upon a successful read, the log flush thread completing;
a reading module of the log read thread configured to read a predefined number of DBMS log records in response to the internal API log having the number of log records satisfying the DBMS threshold, such that the log read thread has an unblocked state;
a notification module of the log read thread configured to notify the log flush thread of a read complete such that the log flush thread completes;
and a bus coupling the processor, storage device, input device, output device, and memory.

15 more elements

Dependencies are inferred from the claim text.