Data Centers

Planning a good backup system

As an administrator, you need to continually monitor and evaluate your backup system. Ron Nutter has an overview of full, incremental, and differential backups, along with suggestions to help you make the best choice for your network.


The only good backup system is one that works and works well. This is one area of your computer network that you need to continually check to make sure that it is working and adjust as needed to keep things on course. Although you won’t see any specific product recommendations in this Daily Feature, I’ll give you an overview of what you need to look at when evaluating your current backup system or when planning a new backup system.

How much to back up and how long you have to do it
One of the first things you will have to decide is how long a backup window you can afford on your network. The backup window refers to the amount of time you can take to back up all of the servers on the network without having to fight users for access to resources. If you are fortunate enough to have an eight-hour window to back up all your servers, you should consider letting the backup job(s) run no more than six hours of the eight-hour window. This gives you some room for growth so that, as more data gets added to the servers, you don’t have to immediately redesign your backup system.

Full, incremental, or differential
There are three types of backups that just about every package will be capable of doing:
  • Full backs up all the files on each selected drive on the servers being backed up and clears the archive bit on all files where it has been set.
  • Incremental backs up only those files whose archive bit has been set and then clears the archive bit on the files that it has backed up.
  • Differential backs up all the files that have changed since the last full backup but doesn’t clear the archive bit.

The differences among the types of backups occur when you have to restore one or more files, or worse yet, an entire server. When restoring from a full backup tape, you will, in most cases, have just one tape to restore from per server. When restoring from an incremental backup, you will have to have at least the last full backup of the server and any incremental backups that have occurred since the last full backup.

If you had to restore a server from scratch or a significant number of files, you would have to restore the tapes in the order that they were made. You would first restore what you needed from the full backup tape and then the incremental backup tapes, in the order they were used. The problem with this type of backup is that if you don’t restore from the tapes in the correct order, you may inadvertently overwrite a newer file with an older file that could be missing some information.

Using a differential backup may be the best of both worlds. Since the archive file attribute flag isn’t reset until the next full backup, in a major restore situation, you will only have to use two tapes for a restore process—the last full backup tape that was used and the most recent differential backup tape.

There will be a slight time difference in backup when you are running the differential vs. the incremental backup. The time required for a differential backup can get progressively longer each night after the full backup, since potentially more files will have been changed since the previous backup was run.

To decide which type of backup to use (you may use one or more of the types listed), you may need to think about what type of backup you will run on a particular day. Since your backup window may be longer on a weekend evening, you may need to restrict the use of a full backup to weekend evenings only.

Keeping the tape drives clean
You should keep the tape drives clean of flakes that break off the tape media on a regular basis. Also, make sure to clean all your tape drives. Check the manual that comes with each tape drive for the recommended cleaning interval.

Some tape drives, such as the digital linear tape (DLT) drives, will have a light on the front panel of the drive that comes on when it is time to clean the drive. After a while, you will know when the drives need to be cleaned even before you see the Clean Tape Drive light for a particular drive. Although you can save a little money by purchasing cheaper cleaning supplies for your backup drive, this is one area that you probably shouldn’t skimp on. Cleaning tapes for a DLT drive can run anywhere from $75 to $100. This is far cheaper than the technician bench time you would be looking at if your local reseller did the job.

Look at the cleaning tapes that the vendor of the tape drive recommends. In the box that held the cleaning tape, you should find a label where you can fill in a square or check a box every time the tape is used. Be sure to check this box each time the tape is used and discard the tape when all the check boxes are filled.

Also check to see what brand of tape cartridge the tape drive manufacturer recommends for use in their drives. Consistent use of the recommended tapes should mean a longer drive life and potentially less oxide flakes falling off the tape cartridge and ending up somewhere inside your precious tape drive.

Choosing the type of backup drive to use
The two most common types of backup drives used to back up servers today are DAT and DLT. DAT (digital audio tape) drives are far cheaper than their DLT counterparts both in terms of the cost of the drive and the cost of the media. A DAT drive capable of backing up 12 gigs noncompressed/24 gigs compressed (assuming a 2-to-1 compression ratio) will run between $1500 and $2000, depending on the manufacturer. The DLT drive that can back up 35 gigs noncompressed/70 gigs compressed will be $3500 or more. Although the cost of the DAT tapes will be less than $20 apiece, DLT tapes will be $80 to $100 apiece, depending on how many you buy at a time.

These prices are for a single drive. When you are backing up just a server or two, a single backup drive will be more than sufficient. When you start backing up 15 servers, though, you will need to look at some type of mini-library or autochanger system that incorporates one or two drives and contains a magazine capable of holding eight to 10 tapes.

When you approach this type of backup system, you may need to rethink how you back up your servers. To keep from having a tape failure prevent you from restoring files to a server because part of the backup is on a tape that is no longer useable, consider creating multiple backup jobs that back up one or two servers each so that no one server’s backup session is spread over multiple tapes.

You also have the option of considering a RAIT tape system. This is similar to RAID hard drives but uses multiple tape drives rather than multiple drives. Like RAID, which provides redundancy for hard drives, in the event a RAIT tape drive fails, you can still restore needed information from the remaining tapes in the backup set.

Conclusion
Maintaining proper backups is an extremely important part of your job. As you can see, this is something you must review on a periodic basis to make sure your backup sessions aren’t running too long and you have a system that is meeting your needs.

Ronald Nutter is a senior systems engineer in Lexington, KY. He's an MCSE, a Novell Master CNE, and a Compaq ASE. Ron has worked with networks ranging in size from single servers to multiserver/multi-OS setups, including NetWare, Windows NT, AS/400, 3090, and UNIX. He's also the help desk editor for Network World. If you’d like to contact Ron, send him an e-mail. (Because of the large volume of e-mail that he receives, it's impossible for him to respond to every message. However, he does read them all.)

The authors and editors have taken care in preparation of the content contained herein but make no expressed or implied warranty of any kind and assume no responsibility for errors or omissions. No liability is assumed for any damages. Always have a verified backup before making any changes.

Editor's Picks

Free Newsletters, In your Inbox