Windows Server 2003 : Planning a Backup Strategy

Understanding Network Backups

A network backup solution consists of three elements: one or more backup drives, a backup software product, and a backup plan that details the use of the other two items.

Backup Hardware

The storage medium that most administrators choose for backing up their networks is magnetic tape. Magnetic tape drives have high capacities, low media costs, and are a reliable means of long-term data storage. High capacity is a major requirement for a backup medium, because administrators usually like to create unattended backup solutions that can run at night, or while the business is closed. The higher the capacity of the storage device, the fewer times an administrator has to change the medium to complete the backup.

Many magnetic tape formats are suitable for backups. Table 1 lists some of these formats. The general rule of magnetic tape drives is that you trade speed and capacity for cost. The best drives on the market can hold up to 200 gigabytes on a single tape, and write data at speeds of nearly 60 megabytes per second. For a drive like this, however, you can easily pay $5,000 or more. Slower drives with lower media capacities can be much less expensive.

For organizations with an enormous amount of data to back up, there are also devices on the market called autochangers or tape libraries. An autochanger combines one or more magnetic tape drives with a robotic mechanism that inserts tapes into drives and removes them. Autochangers range from small desktop devices with one drive that hold a handful of tapes, to huge units the size of a refrigerator, with several drives and holding hundreds of tapes. The advantage of an autochanger is that you can create a single unattended backup job that spans multiple tapes. When the first tape is full, the autochanger removes it from the drive and inserts a new one, repeating the sequence until the job finishes. With a large autochanger, network administrators can sometimes go for weeks or months without having to swap tapes in and out of the device, while their backups proceed automatically every night.

A backup software product is an application that enables you to select the files you want to back up and sends them to the backup drive. Most backup software also includes features that enable you to create repeating backup jobs and schedule them to occur regularly. The usual objective of a network administrator is to create a backup solution that requires as little intervention as possible. With a properly configured hardware and software combination, daily backups should occur with the administrator doing nothing but swapping tapes in and out of the drive.

The backup software must also provide a way to restore data from the backup tapes. When the software backs up data, it creates a catalog of the files it processes, so that you can locate specific data for restoration. You can usually choose individual files for restoration, or restore entire drives, in the case of a disaster.

Network backup software typically differs in several important ways from a software product designed for a standalone computer. Some of these differences are as follows:

• Backup scheduling Some rudimentary backup software products only enable you to perform a backup in real time. Network backup software enables you to schedule backup jobs to occur at any time, and to repeat at regular intervals.

• Remote backup agents Virtually all backup software products can back up a shared network drive mapped to a drive letter on the server, but a remote agent enables the server to back up the entire remote computer, including system state information, such as the registry and Active Directory databases.

• Backup file database Some higher-end network backup software products store the catalogs of the backed up data in a database on the server, providing various ways for administrators to search for files to be restored and to create reports about the data stored on the backup media. Lower-end products store the catalogs on the individual tapes, which can make it difficult to locate the tape containing the specific files you need to restore.

• Media rotation schemes An efficient backup strategy uses a specified number of tapes to back up the network, and rotates them so that the drive overwrites them at regular intervals. However, it is vital that you reuse only tapes that are outdated, and that you don’t overwrite data that you might someday need to restore. Network backup software products typically include a media rotation scheme that helps you schedule your backup jobs and tells you which tape to insert each day.

• Open file backup A backup software product usually cannot back up a file that a running process has locked open. For example, if a user leaves a document open in an application before leaving for the day, the backup software will usually fail to copy that file, because the application has it open. Many network backup products include an open file option that enables the software to back up certain types of files while they are in use.

• Disaster recovery When a computer’s primary hard drive fails completely, you cannot restore the data from your backup right away, because the computer must be able to run the backup program or agent. This means that you must reinstall the operating system and the backup software before the restoration can proceed, which can be a time-consuming process. Disaster recovery software circumvents this problem by creating a backup in combination with a boot disk that you can use on the target computer. The boot disk contains only the operating system software needed for the restoration to occur. After booting from the disk, you can perform the restore, and the computer is back to its original condition much more quickly than if you had to install the operating system manually.

• Database backup Database software products have always been a problem when it comes to backups, because users perpetually leave the databases themselves open. Many network backup software products provide agents that make it possible to back up specific types of databases while they are operational. The agent typically works by intercepting all database access requests and redirecting them to a temporary file called a delta file. While this is occurring, the agent can close the database files themselves and back them up. Once the backup is completed, the agent reopens the database files, and applies any changes that are queued up in the delta file.

Creating a Backup Plan

Once the hardware and software components for your backup solution are in place, the next step is to create a plan that contains elements such as the following:

• What data will be backed up

• When backups will occur

• Which tapes to use and when

Selecting Backup Targets

The easiest way to perform backups of your network is to simply back up all the data on all your computers every day. However, in most cases, this is not a practical approach for reasons such as the following:

• There is too much data to back up. The hard drives that are typically included in today’s computers hold more than ever, and on a large network, total storage capacity can easily add up to thousands of gigabytes. Unless you want to spend an enormous amount of money on tape drive and autochanger hardware, it would not be possible to back up all the data in every computer every day.

• There is not enough time to perform the backups. Most network administrators schedule network backup jobs to occur at night, or whenever the organization is closed. Backing up during off hours makes it less likely for the backup to skip files because they are locked open, and minimizes the impact of the network traffic generated by remote backup processes. For some organizations, the amount of time available to perform backups (called the backup window) would be insufficient to back up the entire network, unless multiple high-speed drives were used.

• There is too much redundant data. Much of the data stored on a typical computer’s hard drive is static; it does not change every day. Application and operating system files never change, and some document files can go for long periods without users changing them. Backing up files like these every day means saving the same data to tape over and over, which is a waste of time and media.

For these and other reasons, backup software products enable you to be selective about the files you back up. As a rule, you should back up every day only the files that change every day, such as frequently used data files. Files that change less frequently are best served by a weekly, or even a monthly, backup. Some operating system and application files need never be backed up because, in the event of a disaster, you would have to reinstall the operating system and applications from your original distribution disks anyway.

Most backup software products enable you to select backup targets in two ways, by checking files and folders in a directory tree display (see Figure 1) or by using filters. Filters enable you to select the files you want included or excluded from a backup by specifying a combination of factors, including file names, extensions, dates, sizes, and attributes. For example, you can select an entire folder containing your Microsoft Word files for backup, and then exclude all the backup copies that Word automatically creates by applying a filter with the file mask Backup*.*.

Understanding Backup Job Types

To simplify the process of backing up only the necessary files, backup software products enable you to select different types of backup jobs. The three most common types are as follows:

• Full backup Copies all the selected files to the backup medium and resets the archive bits for all the copied files

• Incremental backup Copies only the selected files that have archive bits, and then resets those archive bits

• Differential backup Copies only the selected files that have archive bits without resetting those archive bits

The archive bit is a one-bit flag (called an attribute) on every file, which backup software products use to determine whether that file has changed recently. When you perform a full backup, the software resets all the archive bits for the files it has copied to the backup medium by changing the bit values to 0. Later, whenever an application modifies one of these files, it sets the archive bit for that file by changing its value to 1. The next time you perform an incremental or differential backup, the software checks the archive bits of the files targeted for backup and copies only those with archive bit values of 1. The result is that the incremental or differential job has backed up only the files that have changed. You can still restore all the other files from the last full backup, because they have not changed since then.

The difference between an incremental backup and a differential backup is the way that the software treats the archive bits of the files it has just copied. Incremental jobs reset the archive bits and differential jobs don’t. This means that an incremental job consists of only the files that have changed since the last full or incremental backup. A differential job consists of all the files that have changed since the last full backup. The advantage of using incremental jobs is that they occupy the least amount of storage space and take the least amount of time, because the software only writes each changed file to the backup medium once. Differential jobs take up more storage space and take longer to run because the software backs up all the files that have changed repeatedly during each successive differential job until the next full backup.

The advantage of using differential jobs is that the restoration process is simpler and faster. To restore an entire drive that has been lost, you must first restore the last full backup, and then the incremental or differential backups. With incremental backups, you must perform a separate restore for each incremental backup you performed since the last full backup, to ensure that you are getting the latest version of every changed file. With differentials, you only have to restore the most recent differential backup since the last full backup, because the differential contains all the files that have changed since the full backup in their latest versions.

Scheduling Backup Jobs

Most organizations perform incremental or differential backups daily and a full backup once a week. This arrangement provides a good compromise between protection and the amount of time and media devoted to backups. The ideal situation for a backup administrator is having each daily incremental or differential job fit on a single tape. This enables the administrator to schedule the job to run unattended in the middle of the night or during other off hours, without the need to have someone change media. Once you have created your backup schedule, you can simply insert the correct tape into the drive each day. Full backups might require more than one tape, so someone might have to be there to change media.

Creating a media rotation scheme is also part of backup scheduling. The media rotation scheme enables you to use a specified number of tapes for your entire backup strategy, and tells you which tape to insert in the drive each day. As mentioned earlier, some network backup products have preconfigured media rotation schemes that work everything out for you, but the best schemes are those that you can modify to meet your own needs.

Performing Restores

Obviously, exercising all this care when planning and performing your backups is pointless if you cannot restore the data you have backed up. A good backup software product gives you a lot of flexibility in restoring. The software should provide the following basic options:

• File selection You should be able to select any combination of files, folders, or drives on any tape. Some software products enable you to switch between a media view, which displays the contents of each tape in the library, and a disk view, which displays your backup targets and a list of the multiple versions of each file available on your tapes.

• Restore location You should be able to restore your selected files to their original locations automatically, or specify an alternative location; you should also be able to recreate the original directory tree or dump all the files into a single folder.

• Overwrite options When restoring files to their original locations, you should be able to specify the criteria for overwriting existing files with the same names—based on their dates or using other criteria.

Using Volume Shadow Copy

Volume shadow copy is a Windows Server 2003 feature that maintains a library containing multiple versions of selected files. Although not a replacement for system backups, volume shadow copy enables users to access saved versions of files they have accidentally damaged or deleted. This eliminates one of the most onerous chores of the backup administrator: performing single file restores for users who have inadvertently deleted their own files.

To enable volume shadow copy for a volume on your server, you display the Properties dialog box for the volume and click the Shadow Copies tab (see Figure 2). When you select a volume on the list and click Enable, Windows Server 2003 makes a copy of all the files in shared folders on that volume and stamps the copies with the current date and time. As long as shadow copying is enabled for that volume, Windows Server 2003 continues to make two copies a day of these files and saves them until the amount of space designated for volume shadow copies is full. You can modify both the frequency at which the Windows operating system makes copies and the size of the space used to store the copies.

Only computers running Windows Server 2003, Windows XP, and Windows 2000 (with Service Pack 3 or higher) can access the shadow copies of files on your designated volumes. On Windows XP and Windows 2000 workstations, you must first install the client software that makes this possible. Then, a user can access shadow copies by displaying the Properties dialog box for a file in a shadow volume and clicking the Previous Versions tab (see Figure 3)