HANA BACKUP AND RECOVERY: Multi-streaming Data Backups with Third-Party Backup Tools
SAP HANA uses one channel for data backups, by default. With the introduction of SAP HANA SP11 we have new functionality available to make it possible to considerably speed up the backup time by distributing backup data in parallel to multiple devices by using third party back-up tools.
When multiple channels are used, SAP HANA distributes the data equally across the available channels. All the parts of a multi-streamed backup are approximately the same.
Need and Features for Multi-Streaming Backup Channels: For improved performance Backint can use multiple parallel streams for data backups. If parallel streams have been configured, the individual service backups are distributed across all available streams. Different services always use dedicated backup streams. Backups will only be distributed if they are bigger than 128 GB.
- To configure number of parallel streams, use the parallel_data_backup_backint_channels.
- This feature is available starting HANA SP11.
- The maximum number of channels permitted is 32 for each service.
- Both full and delta backups are supported.
- No downtime required to implement backup multi-streaming.
In one of our scenarios we are using three streams for index server for faster database backup:
The index server backup is distributed across three streams.
Name server and XS engine backups are not distributed among several streams because they have smaller than 128 GB backups size.
Procedure: Below are the steps to implement this functionality in your HANA Landscape:
•Backup Team to set parallelism for the system on the network side.
•DB/OPS team to change the Number of Channels for Multi-streaming in HANA Systems.
Steps performed by DB/OPS Team to change number of channels for multi-streaming:
1.In SAP HANA studio, go to the Configuration tab.
2.Expand global.ini and go to section [backup] .
3.Locate the parameter parallel_data_backup_backint_channels.
This parameter controls the number of channels used for multi-streaming.
4.Change the parameter value.
The default value of parallel_data_backup_backint_channels is 1.
In our example we have used 3 streams for index server for multistrea
5.Set parameter data_backup_buffer_size
global.ini -> section [ backup] -> parameter “data_backup_buffer_size” = <512 * Number of Channels>
Database Performance: Monitor next few backups and backup time to check performance improvement. You will notice backup time improvement by 30-40%.Ensure that increase in number of channels does not have a negative impact on memory consumption. As this activity can be performed online but make sure changes can be made in non-business hours to avoid performance degradation in some situations.
Nice stuff Manpreet.
thanks for info
Could you please throw some light on setting parallelism for the system on the network side.How can this be achieved. What all configuration goes in to achieve this.
Thanks in advance.
Thanks and Regards,
is it valid also fo resotre operation?
I would like to know if the setting is different in a HANA scale-out configuration (3+1) ?
What will be the restore strategy?
If I backup using parallelism 2, say for example, then while restoring the DB to point in time, it will use the same number of sessions?