Ted Saul - Offsite Support Consultant
The following terms related to data location are essential
for the discussion of ILM:
- Online Storage - Data is kept on disk for
immediate access by users. A directory command will show the files and all the
information about the files.
- Near-line Storage - Data is not kept immediately
online but considered shelved and located in a format either on another disk or
on tape within an automatic loader device. To the user, the data will show on a
directory command, but upon access there is a slight delay as the data is
automatically "unshelved" from the near-line device. This is an optional step.
- Off-line Storage - This is the traditional
backup scenario where data is backed up to tape. The tape may then have been
removed from the library and stored in a safe location. Information about the
files backed up on the tape may be kept by manual methods or by using an
application cataloging facility. Users will not see the files in their
directory and a strategy needs to be in place for retrieval of this data.
The term access value may be defined as the weight
data carries and corresponds to how quickly it needs to be accessed by the end
user. Data with a high access value will require an immediate response. Current
patient records, for example, have a high access value to doctors, nurses, and
other medical technicians requiring immediate information during the diagnosis
and treatment of the patient. Storing this data off-line on tape for retrieval
would be unacceptable. The same would be true for data in any real-time
transaction processing environments.
Retrieval times for data with a high access value is
measured in seconds and tenths of seconds. Information with a lower access
value might include statistical data such as weather data to help determine the
record high and low temperature for a particular day. This data probably won't
be considered critical and its retrieval could be scheduled and planned.
Another good example of data that may fall into this category is historical
phone charges by a cell phone provider. This data might be classified as
medium-level access and kept near-line or off-line. Immediate access that is
measured in seconds will not be as critical, allowing for an off-line data
storage strategy. Contrast this with data from the current cycle required for
billing, reporting, and even criminal investigation purposes.
It may also be helpful to sub-categorize high access values
into some strategy such as the following.
- High demand access where the data is required to be available almost immediately. Any delay in response will be noticeable and may adversely affect this business it is supporting.
- Medium demand where a slight delay in retrieval may or may not be noticeable but still tolerable.
- Low access where a request for retrieval may be made to another organization and lag time may be measured in a day or so.
These additional access classes help determine which type of
software to use to manage data at certain points in its life. Data with the
highest access value always needs to be kept online and available. The use of
large disk arrays along with database applications can help to manage this data
and ensure that it is always available when needed. Online data protection such
as RAID strategies and volume shadowing are available to protect the
vulnerability of this data between backups.
How long data is to be retained or its retention
requirement is the second important point to consider in its lifecycle. The
period of time that may be required to retrieve and restore will also fall into
this category. It is always important to keep backups of your data, but a
prioritization of backups should be put into place so it is easy to know what
type of recovery is available. Some data may also reach a point when the
retention/restore value is not as high. This typically happens at the end of
the data's lifecycle.
Retention and restore values consider how long data must be
available in one form or another for auditing purposes and how easily the data
can be restored. As with access value, retention might be sub-categorized as:
Medical data is a prime example of data that has a long
retention period. In most cases, patient health information by law must be
retained for seven years or -- in the case of minors -- until they reach the
age of twenty-one. Medical data may not require immediate access but must be
restored upon request.
- High -- where restores may be required within a few minutes. Backups may be sent to disk rather than tape or to libraries where the tapes can be quickly mounted to restore data.
- Medium -- where restores may take a day to complete. Typically this media is left onsite and stored in some type of secure area.
- Low -- where restores may take a day or two particularly if the data is located at an offsite facility.
A system of date-based backups may be setup and retained
with daily, weekly and monthly backups. A weekly backup may replace a set of
daily backups and a monthly backup may replace a set of weekly backups. Yearly
backups may also be captured for long-term storage. Duplicate copies of
long-term backups may be kept as well for disaster recovery purposes.
When thinking long-term, the media used to store data must
be taken into account as well. Consider asking questions such as:
During an actual disaster is not the ideal time to test your
backup strategy. The process for retrieving media, installing software, and
recovering data both onsite and at DR sites should be regularly tested to
ensure the quickest recovery time possible. It will be important to be able to
identify where each critical piece of data exists in its lifecycle to ensure
that no transactions are lost or overlooked.
- Is the life of the media (tape, etc) in use expected to last for the required life of the data?
- Will the hardware be available to handle your current media in the long-term future?
- Should a plan for migration of data be put into place should advanced technology become available?
In HSM, policies can be set up to manage how long data is kept
and maintained before being shelved to a near-line device. These policies can
do the following:
You will want to set your policies so that data with a
medium access value becomes a candidate for shelving. However, data with high
access values or that for any other reason should not be moved, should be set
with the "no-shelve" bit set. Databases are prime candidates for this
type of setting. Current data will be located in the database and archived out
of the database via an export-type command. This exported data may then become
eligible to be shelved.
- Specify which files to move between primary storage and shelf storage.
- Specify which files are not to be moved from primary storage.
- Set a "high-water mark" on primary storage to
automatically trigger shelving on dormant data to shelf storage. A high-water
mark is a defined percentage of disk space used that, when exceeded, causes
shelving to begin.
- Set a "low-water mark" as a space-recovered goal
to limit the number of files that are moved to shelf storage. A low-water mark
is a defined percentage of disk space used that, when reached, causes
policy-defined shelving to stop.
ABS has two policies that directly affect ILM: the SAVE and the ARCHIVE.
The SAVE policy identifies what files or disks are to be
backed up, the schedule to do so, and specific information about the backup.
This is typically unique data that only occurs once within the configuration.
The ARCHIVE policy defines information about the backup usually found to be
redundant with other backups. The ARCHIVE policy is then associated with one or
more SAVE policies. For example, the catalog where data about each backup is to
be stored is written to a field on the ARCHIVE record. There may be multiple
SAVE policies storing their data to the same catalog. In this case, the one
ARCHIVE policy will be associated with all these SAVE policies.
Between the ARCHIVE and SAVE there are seven policy settings
that may affect ILM.
- Retention - Found on the ARCHIVE, retention is the length of time in days to keep
data in catalogs and available for lookup. For example, retention might be
set to 7, 30, or 365 days. The longer the data is kept, the larger the
catalog will grow. It is important to create the correct system of
catalogs to ensure efficient disk usage. For example, yearly backups
should be kept in a catalog of their own. This data will need to be
retained for a long time period and there is no use in having daily or
monthly backups work around this data. The retention setting is the key in
the ILM environment to ensure that metadata about backups is available as
long as needed.
- Scratch Date - The scratch date is the length of time to keep data on tape. Within
ABS/MDMS, a separate volume database tracks information about tapes. The
scratch date is stored on the volume record and is initially derived from
the retention value. It is possible to manually change the scratch date
found on the volume causing a tape to be retained longer than the
information kept by retention in the catalog. This may be useful for
low-priority-retention-valued volumes where some system of manually
tracking these tapes has been put into place. This can help reduce the
size of the catalogs. ILM will use the scratch date to preserve the data
on tape for its appropriate lifetime. Once the scratch date is reached,
the volume may be set to a free state and the data at risk to be
- Offsite Date - Recorded on the volume record as well, this field defines when to
take the tape to an offsite location. Vaulting is a process that needs to
be set up at each site to ensure that backup data is safe from physical
harm. Moving backup volumes also reduces the threat of a single point of
failure by having backups at multiple locations. The offsite date is the
key to its implementation.
- Onsite Date - Also stored on the volume record, this field defines when to bring
the tape back onsite. Vault management settings, onsite, and offsite
dates, play an important role in ILM and are set after a backup is
completed. This may be done via an epilogue command or by using some
- Consolidation - Stored on the ARCHIVE policy, ABS/MDMS uses this field to determine how
long to make or keep a volume set. A volume set is one or more tapes tied
together using the previous and next pointers. There are three different
modifiers available for consolidation: interval, savesets, and volumes.
The most commonly used is the interval qualifier that tells the number of
days that volume set should be active. Once this date is reached, ABS/MDMS
will retire the volume set and start a new one. ILM needs this setting to
ensure that data is movable offsite in a timely manner and does not tie up
large amounts of data in a volume set.
- Expiration date - Expiration is also on the ARCHIVE and is the date the saved data
expires. Expiration can be used as an alternative to retention.
- Catalog - When implementing ABS/MDMS it is advisable to develop a system of
catalogs to ensure time-efficient cleanups, ability to backup and restore,
and proper amount of disk storage available to store the catalogs.
Depending on the amount of data backed up and the length of the retention,
catalogs can become very large and need to be managed carefully. Catalog
is found on the ARCHIVE and may be used for multiple SAVEs. Catalog become
key in initiating faster restores by allowing for the easy location of
data as well as volumes.
To get started on setting up your site-specific Information
Lifecycle Management strategy, it is a good practice to spend time reviewing
the data on your systems, their security requirements, and current locations.
With large arrays of disk with gigabytes and terabytes of data, it is even more
important to get a handle on what files are on your system.
- Define a list of policies and rules that affect the data on your system such as:
With this information, a corporate backup policy can be
initiated that includes how long data should be retained and the process for
handling the data. Security issues should be addressed in the policy as well as
requirements for scrubbing a tape after use.
- Governmental Policy's (e.g., HIPAA, Sarbanes-Oxley)
- Corporate Mandates
- Application requirements
- Commonsense rules
- Divide data farms into zones:
- System Files - System files may change with one of the following conditions.
ECO installed which could affect drivers or other images on the system data files such as:
Configurations to the system such
as new users, security changes, adding of queues, and DECNET and TCP/IP changes
can affect these files.
- Application Files - Upgrades, updates, and ECOs
for specific application may affect the images belonging to applications. These
images may also affect who has access to what data.
- Data Files - The data for an application that
may be contained in text, binary, database, and other type of files. Define
whether they are high, medium, or low access.
- User Files - Those files kept by individual
users for managing their day-to-day work. These files will vary from system to
system but they should not fall into one of the above categories.
- Apply the policies to each data zone:
Write into the backup policy and apply findings from step
one to those in step two. For example, if dealing with patient information
within the medical community, a retention value of seven years may need to be
applied. Accounting data will have Sarbanes-Oxley requirements applied should
any type of governmental audit take place.
The policy should also include a "bare metal restore"
section in case of catastrophic outages. This will address how often
application and system files need to be backed up. Perhaps ECOs or application
updates will only be permitted after a system backup. Backups of original
distribution binaries should also be covered in the policy to ensure their
- Test recovery procedures:.
As mentioned earlier, an actual disaster recovery is not the
time to test your procedures. Schedule specific times to restore tapes and
rebuild systems to ensure they are valid and accessible.
- Review on a regular basis:
Ensure that processes are kept up to date. When
architectural changes take place review your backup policy to make sure that
everything is covered. If using HSM and/or ABS, you may need to review your
policies for them as well. Make sure all your data is being backed up
appropriately and at the expected time. A formal change process that updates
the backup policy during any software or hardware modification is also a good
In the early years of computing, data was used to accomplish
an organization's work on the computer and, in theory, make operations more
efficient. In today's world, though, security and privacy issues along with the
many levels of governmental control have made data a liability to the
organization. It is imperative that it be controlled and protected. Managing
the information lifecycle can be a time-consuming and tedious process but must
remain a priority. Any file created and stored on a computer whether online,
near-line, or off-line should be in its location by design with its movement
» Send feedback to about this article