IT heroes: Masters of the backup universe

When trying to recover from natural or man-made disasters, nothing makes -heroes" out of support personnel faster than a quick, painless crash recovery. To have any hopes of achieving that -hero" status in your organisation, you need to have a solid data backup strategy in place.

There are three basic types of data backups: normal, incremental, and differential. (See Table A.)

This article will explore these various backup types to help you choose the one that's right for your organisation.

Table A
FigureA
Here is a brief explanation of how normal, incremental, and differential methods differ.

Archive bits

Understanding archive bits can give you a better grasp on the mechanics of data backups. Whenever a file is created or updated, an archive bit is tagged onto the filename's attributes. To see a file's attributes, right-click on a file in Windows Explorer and select Properties. The Attributes section, located towards the bottom of the General tab, has an Archive check box. Data backup methods rely heavily on this archive bit. If the check box is empty, the file will not be archived during the next backup process. If the file is updated, a check will be placed in the box, and the file will be archived during the next backup process.

Full/normal data backup

A "normal" or "full" backup archives all files, including system files, application files, user files, and so forth. All archive bits are then reset or cleared. While this method offers excellent protection, it takes a significant amount of time. Restoring data from a full backup, however, is simpler and faster than many other backup methods because all data is readily available on a single tape or a series of tapes.

Unfortunately, normal backups are not practical for every situation. Most normal backups are done at night during nonbusiness hours, and many large organisations have far too much data to back up in such a short time period. One way to resolve this time constraint, however, is to use multiple tape drives that back up specific portions of the server's hard-disk drive. Tape jukebox systems are ideal for large normal backups. For those situations where there is just too much data and too little time, incremental or differential backups may be the key.

Incremental and differential data backups

Because of their discretionary nature, incremental data backups (IDB) and differential data backups (DDB) can speed up the backup process by only backing up new or changed files. As I mentioned earlier, an archive bit is tagged onto a file whenever it's updated. The backup software reads this bit and determines whether or not to archive the file.

Incremental and differential backups are often used in conjunction with a normal backup. Many organisations will make incremental or differential backups on a nightly basis, with normal backups being made over the weekend. This combined method offers both speed and security.

Incremental data backups (IDB)

Incremental and differential backup methods are very similar, except that after an incremental backup is complete, the file's archive bit is reset. This means that unless a file has changed since the last incremental backup, it will be not be archived. Table B illustrates an incremental backup strategy with a weekly normal backup followed by nightly incremental backups.

Table B
FigureB
Here is an example of an IDB anchored by a normal backup.

Let's say File A was backed up on Monday and hasn't changed since. Because the normal backup performed Monday night clears the file's archive bit, subsequent incremental backups will ignore File A. If the file is lost on Friday, you will need the normal backup tape from Monday to restore File A.

File B on the other hand, was backed up on Monday, but it was updated on Wednesday. Since the file now carries an archive bit, an incremental backup will archive the file onto the Thursday-night tape. If the file is lost on Friday, you only need the Wednesday tape to restore File B.

These two scenarios only deal with losing a specific file. So what happens if the entire server crashes? Unfortunately, this is where incremental backups show their biggest weakness. Since it's doubtful that every file is updated on a daily basis, many files are only archived during the weekly normal backup. You will need the normal backup tape as well as all the subsequent incremental tapes in order to completely restore the system.

Differential data backup (DDB)

To make up for the incremental backup process's restoration shortcoming, organisations can use differential backups. As with the incremental process, differential backups archive only updated files, but the files' archive bit is not reset after the archival process. Table C illustrates a differential backup strategy with a weekly normal backup followed by nightly differential backups.

Table C
FigureC
Here is an example of a DDB anchored by a normal backup.

Using our last example again, let's say File A was backed up on Monday and hasn't changed since. Because the normal backup performed Monday night resets the archive bit, subsequent differential backups will ignore File A. If the file is lost on Friday, you will need the normal backup tape to restore File A.

File B, on the other hand, was backed up on Monday, but it was updated on Wednesday. Since the file now carries an archive bit, a differential backup will archive the file onto the Wednesday-night tape. Since the differential backup process does not reset the archive bit, the file will again be backed up onto the Thursday-night tape. If the file is lost on Friday, you only need the Thursday tape. If the file is lost on Saturday, you still only need the previous night's backup. This saves you from having to know when the file was last updated. No matter when the file was updated, you only need the most recent tape.

Because all updated files since the last normal backup are archived on every subsequent differential backup, restoring the entire server is also much simpler. If the server were to completely fail on Friday, you would need only two tapes: Monday night's normal backup tape and the last, or Thursday night's, differential backup tape.

Other types of backups

Although normal, incremental, and differential are the most common backup methods, two other types do exist: copy and daily backups. Copy backups are performed exactly like normal backups, but they do not reset the file's archive bit. Daily backups rely on a file's time stamp to determine if it needs archiving. Daily backups are popular in mission-critical environments where files tend to be updated constantly, requiring multiple daily backups.

Tape condition and tape alternatives

Tape condition can play a vital role in the success or failure of a backup strategy. Always make sure that your tapes are in good working order. Imagine having to tell your manager or CIO that an entire week's worth of company work has been lost because of a $20 tape.

Also, if you can, enable your backup software's "verify" option. After the files are archived, they will be checked for accuracy. Although this usually lengthens the archival process, you'll sleep easier knowing you've made a good backup.

While magnetic tape is still used for the majority of server backups, CDs and DVDs are growing in popularity. However, the limited storage capacity and absence of jukebox-style recorders makes these mediums impractical for large data backups.

Conclusion

When selecting a backup strategy, consider the time needed to back up vs. the time needed to restore and choose a method that fits your organization's needs. It's also a good idea to store your backup tapes at a location other than your organisation's main office. If a fire, flood, tornado, or other natural disaster hits your main office, your data backups will be safe.

TechRepublic is the online community and information resource for all IT professionals, from support staff to executives. We offer in-depth technical articles written for IT professionals by IT professionals. In addition to articles on everything from Windows to e-mail to fire walls, we offer IT industry analysis, downloads, management tips, discussion forums, and e-newsletters.

©2001 TechRepublic, Inc.

Advertisement

Talkback 0 comments

Latest Videos

Sponsored content

Power Centre - Content from our premier sponsors

Blogs

  • Suzanne Tindal Love me, tender
    Considering how expensive and drawn-out tender processes can be to solve problems that might be very immediate, it's little wonder that the Victorian Police IT department tried to work the tender exemptions system.
  • Array 2009 funding drought rolls on
    For Australian start-ups looking for venture capital, 2009 was a very bad year. 2010 may be no better.
  • Array Can not-so-smart meters help the NBN?
    It was interesting to witness Conroy's recent enthusiasm to spruik the NBN's role in supporting the Smart Grid, Smart City initiative. What a pity that Conroy hadn't yet seen the damning report from the Victorian auditor-general about that state's smart-meter roll-out.
  • More blogs »

Tags

Back to top

Featured