Accelerator-based database recovery

1. A computer-implemented method (CIM) comprising:

linking a selected portion of the data stored in a source database to a set of target database(s) by synchronizing the data of the selected portion with copies of the selected portion in each target database of the set of target database(s);

determining that a crash of the source database has occurred; and

responsive to the determination that a crash has occurred, restoring a latest database state of the source database, with the restoration including:

triggering a remote recovery process, and

responsive to the triggering of the remote recovery process, restoring data of the selected portion in the source database from:

copy(ies) of data entries stored in one, or more, target database(s) of the set of target database(s); and

records from a recovery log file of the source database for data entries that have a later timestamp than corresponding copies of the data entry(ies) as stored in the target database(s) of the set of target database(s);

wherein the set of target database(s) includes multiple databases and is a plurality of target databases;

wherein the triggering of a remote recovery process includes initiating parallel recovery streams from at least two different target databases; and

wherein the restoration includes receiving data by the source database through the parallel recovery streams.

2. The CIM according to claim 1 wherein a load due to the parallel recovery streams is balanced between the plurality of target databases according to a predefined optimization rule.

3. The CIM of claim 1 wherein the restoration further includes:

for each data entry of the recovery log file of the source database, determining a log entry time stamp;

for each copy of a data entry of data included in the selected portion, determining a target entry timestamp; and

comparing the log entry timestamps to respectively corresponding target entry timestamps.

4. A computer program product (CPP) comprising:

a set of storage device(s); and

computer code stored collectively in the set of storage device(s), with the computer code including data and instructions to cause a processor(s) set to perform at least the following operations:

linking a selected portion of the data stored in a source database to a set of target database(s) by synchronizing the data of the selected portion with copies of the selected portion in each target database of the set of target database(s),

determining that a crash of the source database has occurred, and

responsive to the determination that a crash has occurred, restoring a latest database state of the source database, with the restoration including:

triggering a remote recovery process; and

responsive to the triggering of the remote recovery process, restoring data of the selected portion in the source database from:

copy(ies) of data entries stored in one, or more, target database(s) of the set of target database(s), and

records from a recovery log file of the source database for data entries that have a later timestamp than corresponding copies of the data entry(ies) as stored in the target database(s) of the set of target database(s);

wherein the set of target database(s) includes multiple databases and is a plurality of target databases;

wherein the triggering of a remote recovery process includes initiating parallel recovery streams from at least two different target databases; and

wherein the restoration includes receiving data by the source database through the parallel recovery streams.

5. The CPP according to claim 4 wherein a load due to the parallel recovery streams is balanced between the plurality of target databases according to a predefined optimization rule.