Data Storage Management

Data storage is one of the fastest growing and most expensive components of a computing center. Unfortunately, storage does not often get as much attention and planning as other parts of the system. A good storage plan, and an implemented program, should provide an excellent payback.

Storage Management Planning

Data Analysis

A first step in a storage management plan should be an analysis of the data being stored. This analysis should include the volatility, importance, storage costs and other information.

Data Volatility

Data can be categorized as static, dynamic and scratch data.

Static data does not change. Static data could include last year's financial results, released engineering drawings, fingerprints, medical images and legal documents.

Dynamic data is working data and is constantly changing. Examples include financial projections, unreleased engineering drawings, personnel records, patient data and inventory levels.

Scratch data is created in the process of generating dynamic or static data. It is not required once the final data is available.

Data Importance

Data is the most important asset of an organization. What if data were lost or destroyed? What if the data were not available for four hours? For a week? What would be the impact on the organization?

Storage Costs

Storage costs include much more than the cost of hard drives and other storage media. They include the costs of backing up, archiving, restoring and managing the data by users and by administrators, non-productive time recovering lost data or waiting for data that is not readily accessible. They also include the costs and disruptions of adding new capacity. Estimates for managing data on magnetic disk are several dollars per megabyte annually, excluding the initial costs of the disk drive device.

Needs Analysis

Scalability

How fast is the data growing? What implications does this growth have for the organization and for the users?

How long must each type of data (engineering, manufacturing, accounting, human resources) be retained?

What hardware and software will need to be replaced? What will they be replaced with?

What new applications will be installed? What additional requirements will they impose?

Will there be new legal and governmental requirements for retaining data?

Performance

How much data needs to be backed up weekly? daily?

Accessibility

Which data has to be accessible immediately? What are acceptable time limits for accessing data that is not required immediately?

Implementation Tactics

Different storage management approaches should be used for static, dynamic and scratch data:

Static data should be archived. Archiving removes the data and the metadata (pointers to the data) from on-line storage. Archiving provides data protection, reduces unnecessary backups and frees up on-line storage space.

Some static data will need to be accessible. For this situation it is recommended that the data be archived, but that its metadata be kept on-line for easier access. Hierarchical Storage Management (HSM) software typically provides this capability.

Other static data will need to be retained for legal or governmental reasons, with very little likelihood of ever being accessed. This data, and its metadata, should be archived. The archive system should provide keyword searches and other methods for identifying files that do not require the requester to remember the file name and path.

At some point in its lifetime, certain static data will no longer be required. At this point the data and its metadata should be removed from the archives and discarded. Getting rid of obsolete or unneeded information is part of housekeeping.

Dynamic data should be backed up on a regular basis. With backup software the original data and its metadata remain on-line. Versions of a file, other than the latest revision can be available with a properly constructed backup process. When dynamic files become static, they are candidates for archiving.

Scratch data does not need to be backed up or archived.

Match Performance with Needs

If data is the most valuable asset in the organization, reliability should not be compromised.

Performance and cost for storage should be analyzed. Not all data will require the same level of performance for a smooth operation of the business:

Some data, typically dynamic data, will require high performance (immediate access, high transfer rates). This data should be stored on media such as disk arrays that provide the required performance. Dynamic data will probably be a relatively small percentage of total data. For better performance, make sure the magnetic disks never get over 85% full, and periodically defragment the data on the disks.
Static data, which is accessed less frequently, can have a lower performance than dynamic data. On-line storage in optical and tape libraries provides access in a few seconds to a few minutes, at a much lower cost than disk storage.
Static data, which has been archived and has a low probability of being accessed, can be sotred off-line on optical platters or tape cartridges.

Data Protection

Protect the data based on its value. As the level of protection increases, the costs to provide that protection also increase, and access to the data generally takes longer. Higher levels of protection should be provided for critical data, and lesser protection for less important data.

Return to LSC's Home Page.

For more information, send us mail inform@lsci.com.