SGI Data Migration Facility (DMF)
Overview
DMF is software to allow automatic migration of files to and from different types of storage media. This effectively allows us to expand the 30TB capacity of our high performance disk system to well over
100TB in the current configuration by using the large tape library. This process is transparent from the user perspective as they simply see a filesystem and generally don't need to worry about
which media a file is located on. Obviously when accessing a file on tape there can be a slight delay whilst the tape library automatically loads the tape into a drive
and reads the data. All inode information (filename, owner, access time etc) is retained on disk so simply running an ls command
doesn't cause files to be read from tape.
How it works
DMF scans each filesystem everyday and ensures that every new file written to disk has a copy created on tape. For the cluster
home directories and any other filesystem which holds the primary copy of the data, a second copy is created on a separate tape
to guard against media errors. Where data on the filesystem is only classed as a backup then only a single copy is created.
DMF also continually checks how full each filesystem is. When a filesystem is more than 95% full DMF automatically removes data
from disk where there is a copy on tape until the filesystem is 90% full. It also checks that at least 80% of the data on disk has
copies on tape in case there comes a need to remove further data as the disk refills. These thresholds can be configured differently for each filesystem.
DMF client commands
SGI provide a set of client commands which allow you to have some control over moving files between disk and tape. The most useful of these are
dmattr - gives information about the status of a file including whether it is on disk or on tape
dmget - forces file(s) to be read back from tape to disk
dmput - forces file(s) to be written to tape
dmdu - displays the total size of files regardless of their location
Atempo Time Navigator (TINA)
Overview
TINA is a professional backup software product which is available for a wide range of client operating systems. It provides the ability to
backup and restore files with a high degree of control using a simple point and click GUI. The restore capabilities include
the ability to restore different versions of files which change regularly and also to do a full point in time restoration of a system to recover
its state at a specific time. Our current configuration uses a virtual tape library which actually writes data onto a DMF managed filesystem.
The current license allows 200 client systems to be backed up.
TINA management interface
TINA has a management interface which is accessed by running the command tina_adm. This allows clients (or agents as TINA calls them)
to be created, modified and deleted. Each client needs to have at least one backup class defined to specify which files and directories
should be backed up. Backup strategies are then defined, an associated with backup classes, to determine when backups should be run.
There is strict control over which users have access to management functions, which clients they can view/modify and from which host system they
connect from to perform management tasks
TINA user interface
The user interface is launched by running the command tina. This allows users to navigate the directory structure viewing both the current
state and previous states at specific times and to select files for restoration. There is full control over dealing with conflicts such as
attempting to restore a file of which a copy already exists.
|
