By default, restore operations use a single stream. You can use multiple streams so that restore operations complete faster.
If you want to perform restore operations to distributed file systems or to network file systems (such as NFS and CIFS) that are mounted across multiple computers, you can use multiple nodes.
For faster multi-stream restore operations, you can define alternate data paths. For more information, see GridStor® (Alternate Data Paths) - Overview.
Components That Multi-Stream Restore Operations Use
Multi-stream restore operations use the following components:
Master node: The destination client that you select for the restore operation is configured as the master node. The master node identifies and coordinates the proxy nodes that are configured for the restore operation. The master node sends the job progress updates to the Job Manager.
Nodes: The data access nodes that you select for the restore operation are configured as proxy nodes for the restore operation. The proxy nodes report to the master node. A controller process runs on each proxy node.
Worker thread: The worker thread is a single-entity thread that is created by the controller process and that actually performs the restore task. Multiple worker threads can run on a proxy node. The number of worker threads that are created must be equal to or less than the number of streams that are configured for the restore operation.
The Multi-Stream Restore Operation Process
During multi-stream restore operations, the Commvault software uses an algorithm to split the data into tasks. For more information, see Algorithm for Splitting Tasks During a Multi-Stream Restore Operation.
After the data is split into tasks, the restore job selects the proxy nodes to perform a parallel, distributed restore operation, considering the highest precedence for a LAN-free restore operation.
-
If the number of proxy nodes that are configured for the restore job is greater than the number of tasks, then all proxy nodes are not used.
-
If the number of tasks is greater than the number of proxy nodes that are configured for the restore job, then one or more proxy nodes run more than one stream.
-
If the number of tasks is less than the number of streams that are configured for the restore job, then all streams are not used.
When the task is complete, the proxy nodes send a Task Complete message to the master node, and the master node updates the Job Manager. If any of the Controller processes or tasks fail, then the master node is informed of the failure. The master node reassigns the pending tasks to controller processes that are capable of handling the tasks. When a task fails, if the task setting specifies, then the restore operation ignores the failure. However, if a hard error causes the task to fail, then the restore operation fails.
Requirements for Multi-Stream Restore Operations
To perform multi-stream restore operations, your system must meet the following requirements:
-
The source client computers and the destination client computers must have V11 Service Pack 6, or a more recent version.
-
The client computer whose data you want to restore must be backed up using Indexing Version 2. For more information about Indexing Version 2, see Indexing Version 2: Overview.
-
For multi-node restore operations, the destination volume must be shared with all the restore nodes and must be accessible from all the nodes.
Agents That Support Multi-Stream Restore Operations
The following agents support multi-stream restore operations:
-
Enhancements in Restores for Guest Files and Folders (supports up to 10 streams, no support for multiple nodes)