User's Manual

132 Migration, release, recall, and deletion
The number of FSE media pools assigned to the same FSE partition equals the number of copies the data
from that FSE partition will have on FSE media. If you have more copies of file data on FSE media, you are
able to recall your data from any available copy, thus benefiting on recall speed and data safety.
Depending on the configuration of your FSE implementation and availability of required resources, you
can have copies made in parallel or sequentially.
Parallel copying
The number of FSE media pools assigned to an FSE partition determines the number of copies of the
migrated data; you set up parallel copying by configuring at least two drives for the FSE partition (the
MaxNumDrivesMigration variable in the FSE partition configuration file). Note that a media pool
cannot be assigned to multiple partitions.
For best performance, there must be at least n drives free when the migration job with parallel copying
starts, where n is the number of media pools assigned to the partition. Because of that, the number of
parallel copies must not exceed the total number of drives that can be used for migration with the partition
(MaxNumDrivesMigration), neither can it exceed the system-wide number of drives that can be used
for migrations (SystemMaxNumDrivesMigration in the FSE system configuration file). Furthermore, the
number of parallel copies must be balanced with the need for drives for other jobs running in FSE.
If there are not enough free drives available during a migration job with multiple copies, the copies that
cannot be made in parallel will be made sequentially.
Sequential copying
The number of configured FSE media pools determines the number of data copies for the corresponding
FSE partition; if you have only one drive configured in FSE or if your FSE partition can use a maximum of
one drive for migration at a time (MaxNumDrivesMigration variable in the FSE partition configuration
file), your copies will be done sequentially. Also, if there are not enough free drives available during the
migration job, copies that could not be made in parallel will be made sequentially. Note that a media pool
cannot be assigned to multiple partitions.
Once a job is started it must be finished as soon as possible to free the disk buffer (the same disk buffer is
used for creating each copy). The job priority increases over time to ensure that the job will allocate
enough resources - drives and media - to complete. If some copies have already been made while others
still need to be made and another job with a higher job priority was triggered in between, the priority of
the migration job is recalculated (increased) using a special factor (PhaseFactor variable in the FSE
system configuration file). Such handling gives the job a high enough priority to allocate the required
resources as soon as possible.
HSM file system access modes
Normally, data on HSM file systems is only available when FSE processes are running, that is, when the
FSE system is in a fully operational state. In certain situations, however, you may need access to the data
even though the FSE implementation cannot be started for some reason. The Limited Access Mode (LAM)
feature of the FSE implementation allows this by enabling read-only access to directories and files on the
mounted HSM file system. While in the read-only mode, the HSM file system is not controlled by
Hierarchical Storage Manager.
NOTE: Limited Access Mode is available on supported Windows operating systems.
On Windows systems, LAM is available with full functionality, including the ability to manually switch
between LAM and FAM and vice versa.
Limited Access Mode (LAM) and Full Access Mode (FAM)
A properly configured HSM file system on a Windows FSE client, from which NTFS has been detached
using the fse --dismount-ntfs can be mounted implicitly or explicitly. The implicit mount is triggered
automatically after an object on the HSM file system is accessed for the first time. The explicit mount is
executed manually invoking the fse --mount command for the corresponding disk volume, and can be
performed only while FSE processes are running.