Improving recovery time

The dsmc restore command starts a single restore job on a single node. This job might need a long period to restore any object data. To improve the restore performance, start separate restore jobs on different IBM Storage Protect client nodes.

You can create separate restore jobs by splitting a single restore task into several smaller ones. One way to create separate restore jobs is to specify the restore path for the object data that is deeper in the IBM Storage Scale object path.

For example, instead of starting the restore with the root of the IBM Storage Scale object path, start the object restore at the virtual devices level. If you have 40 virtual devices that are configured, you might start 40 independent restore jobs to restore the object data. Then, distribute the jobs to the different IBM Storage Protect client nodes. Additionally, you start a single restore job for any files under the account and container path.

With this approach, care needs to be taken not to overload the IBM Storage Protect client nodes or the IBM Storage Protect server. You might want to experiment to determine the most optimal mix of jobs.

For example, if there are four IBM Storage Scale object nodes, each with the IBM Storage Protect client installed and configured, you might use the following types of commands:
  1. On the first IBM Storage Scale object node, run a restore job for each of the first 10 virtual devices by running the following commands:
    dsmc restore /gpfs0/objectfs/o/z1device0 -subdir=yes -disablenqr=no \
-servername=tsm1
dsmc restore /gpfs0/objectfs/o/z1device1 -subdir=yes -disablenqr=no \
-servername=tsm1
    
#<repeat for z1device2 - z1device9>
  2. On the second node, run a restore job for each of the next 10 virtual devices. Continue the pattern on the remaining IBM Storage Scale object nodes so that any virtual devices under the o subdirectory are restored. Also, start a single restore job for any account and container data under the ac subdirectory by running the following command:

    dsmc restore /gpfs0/objectfs/ac -subdir=yes -disablenqr=no -servername=tsm1

The most efficient restore approach depends on many factors, including the number of tape drives, IBM Storage Protect client configuration, and network bandwidth. You might need to experiment with your configuration to determine the most optimal restore strategy.