Auxiliary Copy - Best Practices
Enabling Dynamic Stream Allocation
It is recommended to enable Distribute data evenly among multiple streams for offline read operations when you want to copy large amount of data to a tape library. Few other scenarios in which you should enable Distribute data evenly among multiple streams for offline read operations are given below:
- Multiple Storage Policies are using same library and auxiliary copy jobs for all the storage policies are scheduled at the same time.
- Some drives from the tape library are offline for long time.
Combining Streams Using Multiplexing
For Auxiliary Copy operations in a tape environment, you can utilize streams and tape resources more effectively by enabling Combine Source Data Streams and Multiplex source streams options on a secondary copy.
Avoiding Media Contention
If many backup and restore jobs are using the same storage policy, it may cause the media contention. This affects the performance of each job. To avoid this situation, before starting an auxiliary copy operation for a storage policy, ensure that the storage policy is not used by any other backup or restore jobs. Perform the following before starting an auxiliary copy operation:
- To Identify the Storage Policy associated with any backup or restore job, view the Storage Policy column in the Job Controller.;
- To view all the jobs scheduled for a storage policy, right-click the storage policy and select View | Schedules.
Reducing Level of Multiplexing
If you are using data multiplexing while creating the Primary copy and if all the data from the Primary copy is not included in a secondary copy, then the auxiliary copy operation will become slower. To improve the performance of the auxiliary copy operation, it is recommended to reduce multiplexing factor for the primary copy.
Avoiding Streams with Partially Copied Data in the Waiting State
When running auxiliary copy operation to all the parallel copies configured in the storage policy, we recommend that the source and the destination MediaAgent should be same to avoid streams with partially copied data in the waiting state.