TRIM Integration with Data Protector

25
Figure 27 illustrates such a strategy and what happens when a disaster occurs. When the disaster
occurs, half-an-hour’s data is lost. To reduce the impact on the system, you could configure the backup
to do a full backup nightly and an incremental backup every hour.
Figure 27: High frequency backup strategy
10 am
7 am
8 am 9 am 11 am 12 pm 1 pm 2 pm 3 pm 4pm
5 pm 6 pm 7 pm
Document Content Index
8 pm
Lost
DISASTER
Lost
Lost
Document Store
Meta Data Database
Benefits of this strategy:
• The worst case is the loss of data of one hour.
• There is an extremely low risk of data inconsistency (see ”Procedures Following an Inconsistent Data
Recovery” below for more information on recovering consistency).
High Performance/Low Impact Setup Using Low Frequency Backups
In this scenario we assume that backup to disk, large network bandwidth or servers capable of large
traffic volumes are not available.
To keep the impact on the system as low as possible, consider creating separate backup
specifications and schedules for the individual components of TRIM. Depending on the TRIM setup, it
may be advantageous to back up the Metadata Database more often than the Document Store. For
example, the Document Store could be too large to back up daily, while the Metadata Database is
too important not to back it up daily. This can result in inconsistency, with a Metadata Database more
recent than the Document Store after restoration. However, TRIM can easily handle such inconsistency
as explained in section “Procedures Following an Inconsistent Data Recovery”.
It may also sometimes be sensible to back up the Document Content Index even less frequently than
the Document Store. This can be handled by TRIM very well, since it is possible to recreate the
Document Content Index from the Document Store for a certain time span only. This approach
combines the advantages of not having to back up the Document Content Index too often, and
keeping the recreation time to a minimum.
Figure 28 shows an example how such a backup strategy could be set up. In this example, the very
important Metadata Database is backed up daily, the Document Store is backed up twice a week,
and the Document Content Index only once a week. If a disaster happens as illustrated, the database
loses only half a day, and the Document Store loses one and a half days. Although the Document
Content Index loses four and a half days, it can quickly be recreated from the Document Store so
effectively it also only misses one and a half days of data. The recreation of the Document Content
Index is quick because it only needs to be done for three days.