Dell POWEREDGE R710 - Manual (Page 15)

Dell HPC Tiered Storage Solution
Page 13
If the recall for a file in LTS fails, the Archive agent returns I/O error on failure to recall. If the file is
only partially restored, the stub information is preserved. The next read on the file triggers the recall
again.
All file attributes, including Linux permissions, are maintained on the files that have been migrated to
the LTS tier.
The cluster user stores his/her data on primary tier; the user can then force the data to move from
primary to LTS tier. The user is setup with permissions to do so in the CommCell Console. The user can
then select an archive subclient that they have access to and initiate the archive job.
The user can see the file stub in the originating directory structure and can retrieve the file when
required through the use of persistent recovery.
If the user accidently deletes a file that has been migrated to LTS tier and if a copy of the file remains
on LTS, the user can utilize the end user web search console on the Web Server installed on IIS server
to view and recover their specific archived data back to the original or a staging location.
Primary to scratch to LTS tier
The user can manually or via script, copy/move data from Primary tier /home/user1/jobX, to Scratch
tier /scratch/user1/jobX, as part of the job; and manually or via script copy/move job results back
from /scratch/user1/jobX to /home/user1/jobX. The user can integrate data copy (across scratch and
primary) as part of the job through the use of either the manual operation or the custom scripted
operation.
Performance and scaling
This section describes the performance results of job process times for HSM migration and persistent
data recall events. The test methodology utilizes data sets of various sizes as well as the number of
file(s) in job, to illustrate the effect of overall data set size vs. the number of files within a migration
job. Dell explored data compression states to illustrate performance gains achieved from compression
settings of a 2:1 ratio when compared to no/zero data compression settings (native). In each of the
migration and recall jobs, a storage policy was used to trigger job start and was also configured to
perform data verification to assure data integrity of migrated data.
(5)
The first test, summarized in Table 5.1, tested a single file of various sizes to measure how long it took
to migrate the data to the LTS and leave behind a file stub as well as verify the transaction.