Parallel Replicat is one of the new features introduced in Oracle ® GoldenGate 12c Release 3 (12.3.0.1). Parallel Replicat is designed to help users to quickly load data into their environments by using multiple parallel mappers and threads.
What is Parallel Replicat?
Parallel Replicat is a highly scalable apply engine for the Oracle database that can automatically parallelize the apply workload and take dependencies between transactions into account. Parallel Replicat provides all the benefits of Integrated Replicat by performing the dependency computation and parallelism outside of the database. It parallelizes the reading and mapping of trail files and provides the ability to apply large transactions quickly. The dependency computation, parallelism of the mapping, and apply are performed outside of the database and can be off-loaded to another server. The transaction integrity is maintained in this process.
In addition, the SPLIT_TRANS_RECS parameter splits a larger transaction
into logically smaller pieces to apply in parallel. Dependencies are managed and
maintained as well.
Parallel replication architecture
The following image illustrates the Parallel Replicat architecture:
List cell command results
Image source: https://bit.ly/2wsPZVv
The architecture starts by reading a single trail file but provides a wider road with multiple lanes for reading (mappers) and writing (appliers). Parallel Replicat ensures that all the transactions are ordered based on the key dependencies (primary key (PK), foreign key (FK) and unique key (UK)). This is a huge difference from Integrated Replicat, where the dependency and writing are done within the database.
Additionally, Parallel Replicat can be configured to run in one of the following modes:
-
Integrated mode: This is similar to the prior version’s Integrated Replicat except that the readers and writers are not external to the database with the integrated mode of Parallel Replicat. This mode still uses the internals of the database to manage the processes.
-
Non-integrated mode: In this mode, the Replicat still runs in parallel. However, now it is completely outside of the database.
Parallel Replicat key features
The following features are key for Parallel Replicat:
- Is up to 5 times faster than Integrated Replicat.
- Provides the option to apply a single large transaction in parallel.
- Can parallelize a single large transaction. Processes large transactions faster in parallel and still considers dependencies while paralleling large transactions.
- Can control processing by using the
SPLIT_TRANS_RECSparameter, which specifies the transaction split size (in records). The default is 100,000.
Basic Parallel Replicat parameters
You can use the following parameters can be used in Parallel Replicat processing:
MAP_PARALLELISM: Configures the number of mappers. This parameter controls
the number of threads used for reading the trail file. The minimum value is
1, maximum value is 100, and the default value is 2.
APPLY_PARALLELISM: Configures the number of appliers. This parameter
controls the number of connections in the target database that are used for
applying the changes. The default value is 4.
MIN_APPLY_PARALLELISM and MAX_APPLY_PARALLELISM: The Apply parallelism
function is auto-tuned. You can set a minimum and maximum value to define the
ranges in which the Replicat automatically adjusts its parallelism. There are
no defaults. Do not use this parameter at the same time as the
APPLY_PARALLELISM parameter.
SPLIT_TRANS_REC: Specifies that large transactions should be broken into
pieces of a specific size and applied in parallel. Dependencies between pieces
are still honored. This parameter is disabled by default.
Add non-integrated Parallel Replicat with the adminclient
Perform the following steps to add non-integrated Parallel Replicat with the adminclient:
-
Enter the following commands to open adminclient:
$ cd $OGG_HOME/bin $ bin> ./adminclient -
Enter the following command to connect to the Service Manager deployment source:
adminclient> connect https://<host>:<port> deployment <deploment> as <security user> password <password> -
Enter the following command to create the Parallel Replicat process:
adminclient> add replicat <group name>, integrated, parallel, exttrail <trail name> checkpointtable ggadmin.ggcheckpoint -
Enter the following command to edit the parameter file:
adminclient> edit params <replicat name> -
Enter the following command to start the Parallel Replicat process:
adminclient> start replicat <replicat name>
After the Replicat is created, it automatically shows up in the associated administration service.
After the Replicat starts, the number of threads that you specified for readers (mappers) and writers (appliers) are shown in the report files.
Sample parameter file
The following code is a sample parameter file for Parallel Replicat:
replicat REP1
userid ggadmin, password ****
INSERTUPDATES
REPERROR(1, DISCARD)
MAP_PARALLELISM 2
MIN_APPLY_PARALLELISM 2
MAX_APPLY_PARALLELISM 8
SPLIT_TRANS_RECS 100
MAP *.*, TARGET *.*;
Conclusion
GoldenGate is already a great replication tool from Oracle that provides heterogeneous replication between different types of databases or platforms. Oracle has added an extra advantage to the GoldenGate technology with Parallel Replicat.
Parallel Replicat is a new variant of Replicat that applies transactions in parallel to improve system performance. Parallel Replicat provides all the benefits of Integrated Replicat while performing the dependency computation and parallelism outside the database. It reads and maps all trail files in parallel and provides the ability to apply large transactions quickly in Oracle Database versions 11g (11.2.0.4) and above.