In this chapter, you learn
Disk drives fail, taking critical data with them. People accidentally delete the wrong file. Databases become corrupt, sometimes for no apparent reason. As the network administrator, it's your job to protect against these and other causes of data loss by maintaining backup copies of your important data.
Backing up your data to tape or other removable storage media serves these purposes:
In addition to these reasons for backing up data, proper data backup contributes to a network administrator's employment security. Many organizations have a policy of immediately terminating any network administrator responsible for an unrecoverable data loss. Even without such a policy, failure to properly back up data may result in the need of the responsible person to seek alternative employment.
The advent of RAID and other redundancy options has made some LAN administrators more casual about backing up critical data. In reality, RAID addresses only the first of the four preceding reasons for backing up. The other three reasons make backing up just as crucial as it was before RAID subsystems became a common feature of servers running Windows NT.
Windows NT Server 4.0 provides built-in support for various backup devices, primarily tape backup drives. The Microsoft Hardware Compatibility List for Windows NT, current when this book was written, identifies 145 individual makes and models of backup devices that are compatible with Windows NT Server 4.0. Before you choose backup hardware and software for your Windows NT Server 4.0 installation, however, it's important to understand how Windows NT Server 4.0 handles the file backup process. The following sections explain the purpose of the file archive bit and the types of backup operations supported by Windows NT Server 4.0 and third-party device drivers for backup devices.
You can obtain the current version of the Hardware Compatibility List for Windows NT from Microsoft's Web site at http://www.microsoft.com/ntserver/hcl/hclintro.htm or by FTP from ftp.microsoft.com, in the /bussys/winnt/winnt-docs/hcl folder.
To manage a backup strategy, it's essential that your backup software has a way of knowing when a file has been created or modified since the last normal backup was done. One way to do this is to examine the date/time stamp on each file, and compare it with the time that the last backup was done to determine whether the file has changed. This method, used by the backup applet bundled with Windows 95, is simple in concept but unreliable in use. Using date/time stamps to determine the files to back up is unreliable, because many programs alter the contents of a file but don't change the date/time stamp when doing so. As a result, a better method of determining backup currency is needed. Fortunately, such a method has been available since the early days of MS-DOS.
Like MS-DOS, Windows NT stores with each file an attribute called the archive bit. The archive bit is set to on (a value of 1) when a file is created or modified to indicate that the file hasn't been backed up since the last change to it occurred. Because the archive bit is set to on whenever a file has been written to, it provides a completely reliable means of knowing when a file has changed. Windows NT Server 4.0 backup functions, like all modern full-featured backup programs, uses the archive bit to manage backup.
A normal backup copies all selected files, regardless of the state of their archive bits, to the tape drive or other backup media, and then turns off the archive bit on all files that have been copied (see fig. 8.1). Most third-party backup software refers to this process as a full backup.
8.1
A normal or full backup, which copies all selected files to tape and turns off the archive bit.
A normal backup doesn't necessarily copy all files from a particular volume or disk drive, but it may simply copy a file or set of files from a specified folder or folders on the selected volume or disk drive. What determines a normal backup is that all selected files are copied without regard to the state of their archive bit.
Re-creating a failed hard drive from a normal backup set of the entire drive is straightforward. If the system drive has failed, you must replace the drive and reinstall Windows NT Server before proceeding. If the system drive is operable, after replacing the data drive, use the Windows NT Server backup application, NTBACKUP, to do a full restore of the tape to the new drive. Partial restores, such as those of accidentally deleted files, are equally straightforward. NTBACKUP is described later in the section "Using the Windows NT Server 4.0 Backup Application."
A copy backup is identical to a normal backup, except that copy backups skip the final step of resetting all the archive bits on backed up files to off (see fig. 8.2). Most third-party backup software refers to this process as a full copy backup. The resulting backup tape is identical to what would have been created by a normal backup, but the archive bit status of the files on the disk remains unchanged. The main purpose of a copy backup is to allow you to create an archive or off-site backup set without affecting your main backup set's rotation process.
8.2
A copy backup, which copies all selected files to tape but leaves the archive bit unchanged.
Because the contents of a copy backup set are indistinguishable from those of a normal backup set, restore procedures are identical for these two types of backups.
An incremental backup copies to the backup media all selected files that have their archive bit turned on, and then turns the archive bit off for the files that have been copied (see fig. 8.3). The tape from the first incremental backup done after a normal backup contains only those files altered since the last normal backup. Subsequent incremental backup tapes contain only those files that changed since the last incremental backup. After completing each incremental backup, all files have their archive bits turned off, as though a normal backup were done.
8.3
An incremental backup, which copies only changed files to tape and sets the archive bit off.
Re-creating a failed hard drive from incremental backup sets is a bit more involved than using a normal backup set, because each incremental backup tape contains only some of the changed files, and different incremental backup tapes may contain different versions of the same file. To re-create a failed disk drive, you first restore the most recent normal backup set to the replacement disk. Then you restore all incremental backup sets created after the normal backup set, beginning with the earliest, and proceeding sequentially to the latest incremental backup set.
Restoring an accidentally deleted file is a more complex process. To ensure that you get the latest version of the file, you must start by examining the most recent incremental backup set and work backward until you locate the most recent occurrence of the file on an incremental backup set. If the file in question hasn't changed since the last normal backup, you may have to work all the way back to the last normal backup set before you locate the file. Fortunately, most backup software makes this process somewhat easier by allowing you to search backup logs to locate the file so that you can load the proper tape directly.
The incremental backup is best suited to environments where a relatively large number of different files change each day. Because the incremental backup sets the archive bit off after each of these files is backed up, each file is backed up only on the day that it's changed. This may reduce the number of tapes needed for each daily tape set, and also cuts down on the time required for daily partial backups.
A differential backup copies to the backup media all selected files that have their archive bits turned on, but then leaves the archive bits unchanged on the files that have been copied (see fig. 8.4). This means that each differential backup set contains all files changed since the last normal backup. It also means that each differential backup set is larger than the preceding set, because later sets contain all the files previously backed up, plus all files changed since that last backup.
8.4
A differential backup, which copies only changed files to tape and leaves the archive bit unchanged.
Re-creating a failed hard drive using a differential backup set is relatively straightforward. As with an incremental backup set, you begin by restoring the last normal backup set. Because each differential backup set contains all files changed since the last normal backup, however, you need to restore only the most recent differential backup set. Restoring an accidentally deleted file is similarly straightforward. If the file is listed on your most recent differential backup log, restore from the latest differential backup set. Otherwise, restore from the last normal backup.
A daily copy backup copies all selected files that have been modified that day, but leaves the archive bits unchanged on the files that are copied (see fig. 8.5). Like the Windows 95 backup mentioned earlier in the section "The Archive Bit," the daily copy backup uses file date stamps to determine their eligibility for backup, rather than examine the status of the archive bit.
8.5
A daily copy backup, which copies only changed files to tape, based on the file date, and leaves the archive bit unchanged.
Unlike the incremental backup, which copies all files changed since the last normal backup or incremental backup was done, the daily copy backup must be run at least once each day if it's to be successfully used to archive files changed since the last normal backup. If you fail to run the daily copy backup on one particular day, none of the files changed on that day is written to tape until the next normal, copy, incremental, or differential backup is done. Because it ignores the state of the archive bit, the daily copy backup also fails to back up changed files if the file date stamp wasn't altered at the time the file was changed.
A normal backup has the considerable advantages of thoroughness and simplicity. Each normal backup set contains all the selected files on your hard drive. No juggling of tapes is required to locate a particular file. If a problem occurs, simply retrieve the last normal backup tape and do a full or partial restore, as appropriate. The primary problem with normal backups is that they consume large numbers of backup tapes and take a long time to complete.
Doing a proper backup requires that the contents of the server disk be static-that is, that files not be in use and subject to change by users while the backup process proceeds. For many organizations, this means that the best time to do backups is during evening hours and on weekends. Assuming that the server can be taken down for backup at 7 p.m. and must be back in service by 7 a.m., the backup must be completed within 12 hours.
The fastest 4mm DAT tape drives commonly used for Windows NT Server backup can transfer data at a sustained rate of perhaps 30M/minute, or about 1.8G/hour, allowing time to back up a little more than 20G overnight on weekdays. Even the larger 8mm tape drives with automatic tape changers top out at 60M/minute or so, extending the maximum to perhaps 40G or 50G for an overnight backup. If your server disk "farm" is larger than 40G or 50G, if you have multiple servers to back up to a single tape drive, or if your company's hours of operation are longer, you don't have time to complete a normal backup each night.
Although most servers today aren't a size that causes the time required to do a backup to be an insurmountable problem, this situation will change as larger server disk storage subsystems become the norm. Only a few years ago, 1G was considered a large amount of disk storage for a PC server. Today, 10G arrays are common and 50G servers are available. The increasing use of multimedia and document imaging, plus the decreasing cost of disk storage, means that most server disk farms will continue to grow over the next few years. Because tape drive throughput isn't likely to keep pace, you're likely to find, ultimately, that you no longer have time to do a normal backup overnight.
The second reason that it may not be feasible to use only normal backup sets is tape capacity. Unless your IS department or computer room is staffed 24 hours a day, 7 days a week (24[ts]7), overnight backups require either that the size of the backup set not exceed the capacity of a single tape, or that expensive "jukebox" tape changers are used to do unattended backups.
It's for these reasons that the concepts of incremental and differential partial backups were developed. Using either incremental or differential backups with less-frequent normal backups allows only changed files to be backed up routinely, whereas the unchanged bulk of the disk contents are backed up only weekly, or monthly, typically over the course of a weekend. Because only a subset of the full disk is copied to tape each night, the time and tape capacity factors become lesser issues.
The choice between using incremental backup or differential backup for your partial backups depends on how many files are changed, how frequently the files are changed, the size of the files, and how frequently you expect to need to do restores. If most of your files are large and change infrequently, choosing incremental backup minimizes the total amount of data that must be written to tape, because this data is written only once each time the file changes when using incremental backup (instead of each time a backup is done with differential backup). Conversely, if you have many small files that change frequently, the storage economy of incremental backup is likely to be outweighed by the ease of file retrieval with differential backup.
If the daily copy backup has any valid application, it's to run quick, supplementary "snapshot" backups during the course of the working day. If your main backup rotation does a normal backup each day, you can use either incremental or differential backup to make snapshot backups, because the state of the archive bit doesn't matter to the normal backup run each night. If instead your main backup rotation uses either incremental or differential backup on some nights, any partial backup you do during the day must not alter the archive bits of any files. Incremental backups reset the archive bit and are, therefore, unusable for snapshot backups in this environment. Differential backups don't reset the archive bit; however, using differential backups defeats the purpose of a quick snapshot backup, because the process backs up not just the day's work, but all preceding work as well. If your main backup rotation uses partial backups and you need to do snapshot backups during the course of the day, the daily copy backup may be a useful tool, provided that you keep its limitations in mind. Otherwise, don't consider using daily copy backup.
If you're fortunate enough to have a tape drive large enough to do a normal backup on a single tape, and you have the time each night to complete a normal backup, then do so. Using only normal backups makes it much easier to manage the backup process and much less likely that you might accidentally overwrite the wrong tape or otherwise compromise the integrity of your backup sets. If-as is more common-you must depend on less frequent normal backups and daily partial backups, decide whether using incremental backup or differential backup better suits your data; then use the method that best matches your needs. For typical Windows NT servers, a program using normal backups, with differential backups, is the best choice.
Developing a coherent backup strategy that reliably safeguards your data requires more than simply deciding to do a normal backup each weekend and a partial backup every night. You must also consider several other factors, discussed in the following sections, that bear on data integrity and managing the backup process.
The first issue to consider when developing a backup strategy is how you arrange the data on your server's fixed-disk drives. If you have only a single disk volume on your server, there's nothing to decide. If, however, you have multiple volumes, you can decide what type of data resides on each volume. Figure 8.6 shows one possible arrangement of data on volumes intended to make backup easier. Making the correct data organization decisions eases the entire backup process; making the wrong decisions can complicate backup operations needlessly.
8.6
Organizing disk volumes to optimize backup operations.
A satisfactory organization places user home directories and other areas with files that frequently change on one or more volumes but segregates system files and other files that change infrequently in a separate volume of their own. Depending on the number of volumes you've created and the types of data you must store, you can extend this file segregation process. For example, if you have a large database that's updated infrequently, you may decide to place it on a dedicated volume, thereby minimizing or obviating entirely the need for backup of that volume. Similarly, if you have large imaging files that change frequently, you may allocate a volume to them and then use incremental backup for only that volume, using differential backup for other volumes where its use is more appropriate. Always manage your volumes with backup issues in the back of your mind.
Client/server relational database management systems (RDBMSs), such as Microsoft SQL Server 6.5, present special difficulties in backing up data because their files remain open whenever the RDBMS is running. Most RDBMSs therefore provide a backup system independent of the backup application supplied with Windows NT Server 4.0.
RDBMS backup procedures use a periodic database dump, which is equivalent to a normal backup. Incremental backups save the content of a transaction log, which lists all modifications to the database since the last dump. On completion of a database dump, the transaction log file is deleted. RDBMS transaction logs vary from the transitory transaction logs created by Windows NT's Log File Service for NTFS volumes.
The second issue to consider when developing a backup strategy is how you verify the integrity of your backups. Many LAN administrators have found, to their sorrow, that the backup tape they relied on to restore a failed drive is unreadable. You don't want to find this out when it's too late to recover your data. The way to avoid this problem is to perform a compare operation to verify that the contents of the backup tape correspond to the contents of the disk drive.
Inexpensive tape drives use a single head to perform both read and write operations. Doing a compare with these tape drives requires a second complete pass of the tape through the drive, doubling backup time. More expensive tape drives have separate read and write heads, accommodating a process called read-after-write. With these drives, the write head records the backup data to tape, which is then read immediately by the second read head and compared with the data stream coming from the fixed-disk drive. Read-after-write tape drives definitely are preferred, because such drives can perform both a backup and compare during a single tape pass, assuming that read-after-write is supported by the backup software.
If you find yourself forced to use a tape drive that requires a second pass to compare files, take the extra time to do the compare. Otherwise, you may find yourself with a failed disk drive and no backup from which to restore your data. Many network administrators with low-cost tape drives compromise between running a compare after every backup and never running a compare operation. A common procedure is to run a compare after each normal backup but run partial backups without doing the compare. Another, somewhat riskier, method is to do the compare on partial backups, where a compare runs more quickly, and not run the compare on the larger, multitape normal backups. Whichever method you choose, make sure that you do a compare at least occasionally, to verify that your tape drive is really writing readable data to your tapes.
Periodically reading backup tapes created by one tape drive on another tape drive that accommodates the same format is a necessity to ensure that you don't have tape interchange problems. If you have an undiscovered interchange problem and your source tape drive fails or is destroyed in a calamity, you can't restore your backup by using another tape drive. Interchange problems usually occur as a result of mechanical changes in the tape path or head alignment that don't affect compare operations on the same drive. Interchange problems are more likely to occur with high-density tape formats, such as 4mm DAT and 8mm.
The third major issue to consider when developing a backup strategy is how to handle the backup of open files. Open files on the network pose a difficult problem for network administrators who want to make sure that all files are successfully backed up. Most backup programs, including the Windows NT Server 4.0 backup application, let you specify how to handle open files on the network. You can elect to bypass open files during the backup operation, or return to open files at the end of a backup session to determine whether the open files have been closed and can therefore now be backed up.
Backing up an open file also isn't practical because the file may be written to and closed while being written to the tape drive. This results in a corrupted file stored to tape. Consider the result of backing up an open database file that has related index files open. The main database file is successfully written to tape in its current state, and the backup software continues to read other files from the disk. A user saves data, modifying both the main database file and one or more index files. (Separate data and index files are used by desktop database systems such as dBASE, FoxPro, and Paradox.) Because the main database file has already been written but the index files haven't, the resulting backup tape contains an older version of the main database file and newer versions of the index files. If you then restore this database and its associated index files, the index files don't match the database files, and applications can't access the database files.
Clearly, it's preferable to perform backups when all files are closed and you're the only user logged on to the server. This may not always be possible, however-particularly if your server must be operational 24 hours a day, 7 days a week. Many firms that operate Internet servers don't want to discourage connection to their Web site by taking the server out of service for backups.
The only way to handle the open file backup problem is to use a RAID subsystem (described Chapter 7, "Setting Up Redundant Arrays of Inexpensive Disks (RAID)") to provide built-in redundancy for your disk storage. Other than the additional cost, RAID has two drawbacks:
Before purchasing a tape drive or other means of backup, carefully consider the size of the largest backup you're likely to need. Ideally, the capacity of the tape drive you select should exceed by a comfortable margin the size of the largest backup data set you make during the life of the tape drive. The goal is to ensure that any backup set, regardless of its size, can be written to a single tape. Avoiding backup sets that span multiple tapes makes backup administration and tape rotation considerably easier and minimizes the chance for using the wrong tape by mistake.
When selecting a tape drive, keep in mind that the disk storage on your server is likely to grow larger as time passes. Remember also that the nominal capacities stated for tape drives assume 2:1 compression of the data being backed up. A tape drive rated at 14G capacity may have only a 7G native capacity, depending on file compression to gain the remaining 7G. If you back up a typical mix of files on a server used primarily for office automation tasks, you may find that the drive achieves the estimated 2:1 compression ratio. If you're backing up a disk drive dedicated to fixed images or video files, which usually are stored in a compressed format, you may find that you achieve compression ratios of 1.1:1 or less.
Tapes should be rotated in an organized manner to satisfy the following five goals:
Various tape-rotation schemes has been developed. Some are quite simple in concept, at the expense of inadequately meeting one or more of the preceding goals. Others are quite complex and meet all the goals for a good rotation scheme, but are difficult to administer on a day-to-day basis.
Some tape-rotation methods use normal backups exclusively; others use less frequent normal backups with daily incremental or differential backups; still others can be used either way. You may be forced to use a rotation that includes partial backups if your tape drive can't store a normal backup set to a single tape, or if time constraints don't allow daily normal backups. If neither condition applies, you're better served by a rotation that uses normal backups exclusively.
The following sections describe five different types of rotation methods. Each method assumes that a normal backup can be done to a single tape and that your organization operates on a five-day-per-week schedule. The methods may be modified if one or both of these conditions can't be met.
The four-tape rotation method, shown in figure 8.7, is very commonly used for small servers. The backup tape set comprises at least four tapes, labeled Weekly A, Weekly B, Daily A, and Daily B. The rotation begins with a normal backup to Weekly A on Friday of Week 1, followed by a daily differential backup to Daily A on Monday through Thursday of Week 2. On Friday of Week 2, a normal backup is done to Weekly B, followed by daily differential backups to Daily B on Monday through Thursday of Week 3. On Friday of Week 3, a normal backup is done to Weekly A, and the cycle begins again.
8.7
A weekly normal backup with daily differential backup using four tapes.
Although the four-tape method is simple to administer and uses a minimum number of tapes, it suffers the following drawbacks:
You can make the following minor modifications to this rotation scheme to address the preceding problems without substantial complication of the backup administration process:
This version of the 10-tape rotation method, shown in figure 8.8, uses 10 tapes, labeled Weekly A, Weekly B, Monday A through Thursday A, and Monday B through Thursday B. The rotation begins with a normal backup to Weekly A on Friday of week 1, followed by a daily incremental backup to the appropriately labeled daily A tape on Monday through Thursday of week 2. On Friday of week 2, a normal backup is done to Weekly B, followed by daily incremental backups to the appropriately labeled daily B tape on Monday through Thursday of week 3. On Friday of week 3, a normal backup is done to Weekly A, and the cycle begins again.
8.8
A weekly normal backup with daily incremental backup using 10 tapes.
This version of the 10-tape method shares the first two drawbacks of the four-tape method, allowing only two weeks of historical data and making no provision for off-site storage. These problems can be addressed by making a copy backup each month, which is stored off site and rotated on a quarterly or annual basis. The third problem, loss of a daily tape, is more difficult to address. Because a daily incremental backup tape contains only a subset of the files changed during that week, loss of one daily incremental backup tape may require you to return to the last normal backup to retrieve a copy of a particular file. Also, if you need to restore a failed hard drive, loss of one incremental backup tape, which happens to contain critical or interdependent files, may force you to roll your restore back to the last preceding normal backup. The only way to protect against this problem is to run the incremental backup twice each day to different daily tapes, defeating the purpose of using a partial backup rotation scheme.
Although this version of the 10-tape method is commonly used, it offers the worst of all worlds. This version of the 10-tape method puts your data at risk, uses no fewer tapes than other methods, and saves very little time compared to using differential backups. Don't consider using this version of the 10-tape method.
This alternate 10-tape method, shown in figure 8.9, is the simplest tape-rotation system in common use that uses normal backups exclusively. It uses 10 tapes, labeled Monday A through Friday A and Monday B through Friday B. A normal backup is done each day to the appropriate tape, and the full set is cycled every two weeks. The major drawback to this method is that it limits historical data to a two-week span. Again, the historical problem can be addressed simply by doing a biweekly or monthly normal backup to a separate tape and archiving the tape. The archive tapes can be rotated on a quarterly or annual basis.
8.9
A daily full backup with two-set rotation using 10 tapes.
The grandfather-father-son (or GFS) method, shown in figure 8.10, is probably the most commonly used tape-rotation method. It's relatively easy to manage, fairly efficient in terms of the number of tapes required, and is supported by almost every backup software package on the market. A GFS rotation can use normal backups exclusively, or can use a combination of normal backups and partial backups.
8.10
A grandfather-father-son tape rotation using 10 tapes.
GFS is nearly always an acceptable, if not optimum, choice for a small- or medium-size Windows NT server. The two primary drawbacks to GFS are that it uses some tapes more heavily than others and that, in its unmodified form, makes no provision for balancing off-site storage needs with quick retrieval requirements.
A typical GFS rotation scheme requires 21 tapes to be used over the course of a year, although this number can be altered depending on your particular archiving needs. Four daily tapes are labeled Monday through Thursday. Five weekly tapes are labeled Friday 1 through Friday 5 (the fifth Friday tape accommodates months with five Fridays). Twelve monthly tapes are labeled January through December.
The daily tapes are used on the day corresponding to the tape label, and are overwritten every week. Each Friday, the correspondingly numbered Friday tape is used, meaning that Friday tapes are overwritten only once each month. On the last day of each month, a normal backup is done to the corresponding monthly tape, which is therefore overwritten only once per year.
Depending on your needs, you can alter the time span between archival backups. Firms for which archiving is less critical can substitute quarterly tapes for monthly ones. Firms for which archiving is very important may choose to substitute biweekly or even weekly archive tapes for the monthly tapes described here.
The Tower of Hanoi (TOH) rotation method, shown in figure 8.11, is named for a game that uses three posts and several rings of various diameters. The object of the game is to relocate the rings by using a minimum number of moves so that the rings are placed in sequence on a post, with the largest ring on the bottom and the smallest on the top. The TOH backup model introduces new tapes to the backup set periodically, using the newly introduced tape every other rotation. TOH doubles the time before the previous tape comes back into the rotation, consequently doubling the time before earlier tapes are introduced back into the rotation.
8.11
The Tower of Hanoi tape rotation.
The TOH rotation has many advantages, but at the cost of considerable complexity in terms of managing the tape rotation. Wear on individual tapes is distributed relatively evenly over time using this method, but the real advantage of TOH versus GFS and other methods is that TOH saves many versions of each file, allowing you to selectively retrieve different versions of the file.
The TOH rotation originated for mainframe and minicomputer backup, and has since migrated to the PC LAN environment. For a long time, Palindrome's PC Backup software was the only product supporting this rotation, although TOH rotation is now becoming more common. Backup software from Symantec and other third parties offer automated TOH implementation.
If the backup software you select supports a Tower of Hanoi rotation, consider using it for its many advantages, particularly if yours is a medium or large LAN. If your software doesn't provide TOH, don't even consider an attempt to implement TOH rotation manually. You will almost certainly use the wrong tape at one time or another, destroying the integrity of the rotation.
A concept called Hierarchical Storage Management (HSM) was originally developed in the mainframe arena and is beginning to migrate to PC LANs. HSM is based on the fact that, although fixed-disk drives offer fast access to data, they are an expensive commodity of finite capacity on real-world LANs. Tape and other backup media offer much slower access, but can be extended to essentially unlimited capacity simply by adding more tapes. HSM uses fast but limited-capacity fixed-disk drives to store backup data that's needed quickly, plus tapes to store an unlimited amount of data that doesn't need to be retrieved as quickly.
HSM categorizes storage devices as online, near-line, and offline:
The key to HSM systems is that files are automatically migrated from the faster (and more expensive) online storage devices to near-line storage and eventually to offline storage, based on the usage patterns of the particular file. This process is managed by the HSM software itself, without the need for routine intervention by the LAN administrator. Infrequently used files, which must nevertheless remain available, can be designated as near-line.
HSM has become a hot topic during the last 18 months or so, and HSM-based products for Windows NT Server are shipping from (or have been announced by) most major backup software suppliers, including Seagate/Palindrome, Conner/Arcada, Cheyenne, and others. If your storage requirements can benefit from HSM, examine the capabilities and costs of these products before deciding on a final backup solution.
Simple backup solutions, both hardware and software, focus on the needs of relatively small server backup in a single-server environment. If your LAN includes multiple servers, multiple sites, huge amounts of server disk storage, or the need to back up user workstations running diverse operating systems, consider acquiring an enterprise backup solution.
Enterprise backup systems combine sophisticated backup software, often including HSM capabilities, with hardware designed to offer an integrated, centralized backup solution for the entire enterprise. Although enterprise backup systems aren't cheap, the alternative is replicated software and hardware on multiple servers.
Palindrome's Backup Director and Legato's Network Archivist products historically have been the leaders in the enterprise backup market, although competing products from Arcada, Cheyenne, Symantec, and others are now beginning to appear. Used with high-capacity tape drives or tape auto-changers, these products automate the backup process for a large multiserver network, allowing administration from a central location.
One unique product deserves mention in the enterprise storage category. The Intel Storage Express system is a turnkey software and hardware solution to enterprise backup. Intel Storage Express handles multiple servers running different network operating systems, including Windows NT Server and Novell NetWare. Storage Express can back up drives of individual workstations running DOS, Windows 3.x, Windows NT, OS/2, and several flavors of UNIX. Storage Express is expandable on a modular basis to provide as much as 144G of tape capacity, using either 4mm DAT or 8mm drives.
The issue of off-site storage of backup tapes presents a conundrum. Although you want to keep your data stored safely off site to guard against catastrophic damage to your LAN, at the same time you want backup copies of your data to be readily accessible when you need them.
There are at least two good reasons to maintain an off-site set of backup tapes. Most obviously, doing so protects against catastrophic data loss due to fires and natural disasters, which may claim your on-site backup tapes at the same time they destroy your server.
Less obvious but just as important to many companies is the need to maintain archival sets of backup data for legal or other reasons. It may be necessary to keep several years' worth of data relating to tax issues, personnel files, and so on. Space constraints alone often make it necessary to keep these archives off site.
The best available fire-safe storage units can't guarantee that fragile tapes will survive a major fire. Many businesses suffering a catastrophic data loss find themselves out of business shortly thereafter; thus, there's no substitute for off-site storage. It's obviously preferable that the contents of the tapes stored off site be as up-to-date as possible. However, because the off-site tapes are a last-ditch defense that are never likely to be used, some currency sacrifice may be acceptable.
The opposite side of the safety issue is the need for backup data to be quickly accessible. If a hard drive fails on a production LAN, you need to restore as soon as possible, and you need to do the restore from the most recent backup. Having the LAN down can cost your company hundreds, thousands, or even tens of thousands of dollars per hour in lost productivity and orders. Having to wait several hours or longer to get your hands on the most recent backup is unacceptable.
How you manage this problem depends on how recent you require your off-site backup to be, versus how quickly the most recent backup must be accessible. Organizations that demand the best of both worlds have no real alternative but to use duplicate normal backups, cycling one copy off site each day and keeping the other available for local use. If your requirements are less demanding, you can simply keep the most recent normal backup on site and cycle the next-most-recent normal backup to off-site storage on a rotating basis. If you require only minimal off-site protection, you can simply perform a monthly archive backup and move it off site.
Regardless of how frequently you rotate backups off site, the issue arises of where to store the tapes. Although various services offer secure storage of off-site backups, most organizations work on a less formal basis. Short of nuclear war, it's very unlikely that both an on-site backup and an off-site backup stored at an unprotected location, such as your home, would be destroyed simultaneously. Giving up the marginal added safety of using a commercial off-site storage facility pays dividends, not only in terms of costs not incurred but in terms of ease of rotating the off-site backup sets. Keep your off-site sets at home, and develop a rotation scheme to keep the tapes refreshed frequently.
It's easy to focus on backing up to the extent that you lose sight of the real purpose of backing up data, which is to restore the data when necessary. The objective is to back up frequently and restore very seldom, if ever. More fundamentally, the goal of backing up is to ensure that if the server crashes, you can get the server back up and running quickly.
Most network administrators back up dozens or hundreds of times for each episode requiring a full restore. Network administrators may perform partial restores of accidentally deleted files on a weekly or even daily basis, but full restores are rare, usually occurring only when the server disk storage system is upgraded, the operating system is changed, or a disk drive crashes. You don't get much practice doing full restores. To make matters worse, when you do need to do a full restore, you're often working under the gun, with a crashed hard drive and many upset users waiting for the server to return to life. Accordingly, it's worthwhile to spend some time to develop a plan to make sure that when you do need to do a full restore, you can do so easily and quickly. Following are the four items necessary to assure a successful data-restore process:
Even carefully planned backup and restore policies and procedures are of little use unless a qualified person is present to implement them when needed. A LAN administrator will sometimes be sick, on vacation, at a conference, or otherwise away when a restore needs to be done.
Make sure that at least one, and preferably several, of your staff members have been trained to follow the backup and restore procedures you've implemented. An untrained person who tries to restore data to your network server may do more harm than good. Also make sure that each of these "pinch hitters" is fully informed of where tapes are stored and can get to them, if needed.
You can do little to make the restore run faster, but one step guaranteed to save time is having a prebuilt C system disk sitting on the shelf, ready to use. This disk should be bootable and should include a full installation of Windows NT Server 4.0, along with all special drivers needed to support the peripherals on your server. It should also include an installed copy of your backup software. If your system drive fails, you can plug this spare drive into the server and avoid wasting an hour or two tracking down distribution disks, reinstalling Windows NT Server 4.0 and its drivers, and getting your backup software installed and running.
Backup and tape drives traditionally are considered synonymous. Although tape is the overwhelming choice of backup media for most network administrators, various optical storage technologies are beginning to nip at its heels. These technologies are still niche products, insofar as the backup market is concerned, both because they're largely proprietary in nature and because their cost per byte stored is still relatively high in most cases. Still, it's worth examining some of these alternatives briefly, both for their current value in fitting specific needs and their possible future value as an alternative to tape. The following sections describe the relative merits of common tape backup formats and alternative optical storage systems for backup and archiving data.
Tape drives are the traditional method for backing up data. In terms of reusability and cost per byte stored, the tape drive is now-and is likely to remain-the best choice for backing up Windows NT 4.0 servers. The following sections describe the most common types of tape drives in use today.
Most tape drives sold today are quarter-inch cartridge or QIC-compatible. The following form factors are used for QIC cartridges:
QIC-80 minicartridges originally stored 80M on a 205-foot tape, but they have since been expanded. 120M capacity is achieved by the use of 307 1/2-foot tapes, and 250M is achieved by the use of data compression. The newer cartridges with 350M capacity use an even longer tape. QIC-80 drives are controlled by the diskette controller and are too small and too slow for serious consideration as a backup solution for all but the smallest servers.
In response to the rapid growth in the size of disk drives, the Quarter-Inch Cartridge Standards Committee approved the QIC-3010 standard in 1991 and the QIC-3020 standard in 1993. The QIC-3010 standard specifies a 3 1/2-inch minicartridge tape with a native capacity of 340M and a compressed capacity of 680M. The QIC-3020 standard specifies a 3 1/2-inch minicartridge tape with a 680M native capacity and a 1.36G compressed capacity.
Sony recently introduced the QIC-Wide cartridge, using 0.315-inch (8mm) wide tape, rather than the 0.250-inch tape used in earlier QIC cartridges. QIC-Wide expands native storage capacity from 120M to 210M, usually advertised as 420M compressed. QIC-Wide drives can read and write both QIC-Wide and standard QIC-80 tapes, but standard drives can't read QIC-Wide tapes.
The QIC-Wide cartridge seems destined to die an early death, due to its marginal increase in capacity and that Sony is the sole source of tapes, which are rather costly compared with standard QIC tapes.
Another scheme to extend the usefulness of the QIC-80 format is 3M's recently introduced DC-2120XL tapes. Usually sold in a kit containing two DC-2120XL tapes and a copy of Arcada Backup 4.1, these tapes increase the native QIC-80 capacity to 175M, again usually advertised as 350M with compression. These tapes have the advantage of working in an unmodified 120/250M QIC-80 drive. (It's interesting that the Arcada backup software included in the kit can't format a tape, thereby presumably ensuring a continuing market for preformatted 3M brand tapes.)
The most recent activity on the QIC-80 front has been the introduction of Travan technology by Hewlett-Packard's Colorado Memory Systems subsidiary, which manufactures the drives, and 3M, which produces the tapes. First shipped in the summer of 1995, Travan drives use tapes that look like a lop-sided QIC-80 minicartridge. The drives, compatible with QIC-80, QIC-3010, and QIC-3020 media, can read from and write to standard QIC-3010 and QIC-3020 cartridges.
The initial Travan media, designated TR-1, provide 400M native capacity, with compression achieving 800M per tape. Travan levels TR-2 and TR-3 began shipping in the fall of 1995, and TR-4 was released in mid-1996. The TR-2 tape uses a modified QIC-3010 cartridge to provide 800M native capacity, yielding 1.6G with compression. The TR-3 tape provides 1.6G native capacity, using a modified QIC-3020 cartridge, yielding a nominal 3.2G of storage using compression. The TR-4 tape stores 4G natively, or 8G with compression.
TR-1, TR-2, and TR-3 cartridges each contain 750 feet of 0.315-inch-wide tape. The TR-1 uses 550-Oersted media to record at 14,700 flux transitions per inch (ftpi), whereas the TR-2 and TR-3 cartridges use higher coercivity 900-Oersted media to record at 22,125 ftpi and 44,250 ftpi, respectively. Because Travan technology is patented by 3M, tapes are available only from 3M and are quite expensive, about $35 each (street price).
These QIC minicartridge drives share two failings, both due to the desire of tape drive manufacturers to supply inexpensive drive mechanisms:
Tape-drive manufacturers realize that with workstations commonly equipped with 1G or larger fixed-disk drives, the slow data transfer rates of QIC drives are becoming unacceptable. The manufacturers have tried to address the problem by introducing accelerator cards that double data transfer rates to 1mbps, the same as that of 2.88M diskette drives. QIC-3020 drives support 2mbps transfer rates with special controller cards. Even at best, 2mbps translates to only about 12M/minute. At this rate, backing up 6G of server disk takes more than eight hours and requires changing tapes at least five times. When you remember that these inexpensive drives also require a second pass to do a compare, it becomes obvious that server backup needs a better solution.
Some newer QIC-3020 drives use the ATAPI IDE interface, so backup speeds will be constrained by the tape-drive mechanism rather than by diskette data transfer rate. Given that media capacities are increasing-particularly with Travan TR-2, TR-3, and TR-4-it remains to be seen whether these faster QIC drives will provide adequate performance and capacity for low-end servers. The drives' lack of separate read and write heads precludes read-after-write, requiring a second compare pass.
The newest technology in tape backup is called Digital Linear Tape, or DLT. Because it's very fast, offers large capacities, and is extremely reliable, DLT is beginning to replace DAT and 8mm tape drives in large server environments. Current DLT tape drives have capacities of 40G compressed, or 20G uncompressed, so even a very large disk subsystem can be backed up to a single tape.
DLT divides the tape into multiple parallel horizontal tracks. While the single write head remains stationary, the DLT drive streams the tape past it, allowing the DLT tape drive to record information to tape as fast as the server can supply it. This removes the tape drive as the bottleneck restricting backup speed and limits backup performance only by the maximum throughput available from the server.
DLT tape drives and their tapes are also extremely robust. A typical DLT drive is rated at 15,000 in-use hours MTBF (mean time between failures). The tapes themselves are rated at 500,000 hours, so the average tape should outlast the technology itself.
Originally developed for use in video recording, helical scanning works by running a tape past a head that rotates at an angle relative to the motion of the tape. The resulting tracks, shown in figure 8.12, resemble diagonal lines running from one edge of the tape to the other, repeating this pattern from end to end on the tape.
8.12
A helical scan tape drive, which uses a head angled relative to the direction of tape movement to record diagonal tracks on the tape.
The advantage of helical scanning is that information can be packed more densely on the media. The most common application of helical scan recording is the consumer VCR; both VHS and 8mm VCRs, as well as the new Digital Video (DV) camcorders and VCRs, use helical scanning. Helical scan recording is a mature technology, although it's more expensive to implement than the linear recording used in QIC-80 drives.
The following helical scan tape backup technologies were available when this book was written:
The best backup solution for most Windows NT Server environments is either 4mm or 8mm tape. Both drive types are available only with a SCSI-2 interface. The drives are fast, inexpensive, and reliable; the tapes are small and inexpensive.
Writable optical drives today are used primarily for archiving data, rather than backup. Various new technologies include erasable CD-ROMs (CD-E) and writable digital video discs (DVDs, also called digital versatile discs), which store up to 4.7G. Even farther in the future are exotic writable disc technologies, based on cholesteric liquid crystals (CLCs), which promise up to 280G of storage per side.
The following sections describe currently available drives that use lasers to write to and read from discs that, for the most part, have the same dimensions as conventional audio CDs and CD-ROMs.
Recordable CD (CD-R) technology has been around for a few years, but it's just now joining the mainstream. With prices on drives dropping below the magic $1,000 point, CD-R drives are poised to take off. CD-R drives are similar to standard CD-ROM drives, but CD-R drives use a higher powered laser that can write to specially designed CDs. These CDs can then be read in any standard CD-ROM drive.
CD-R's relatively low capacity of 680M, its use of relatively expensive media (about $7 per disc), and its lack of rewritability make CD-R a poor choice for routine backup. The first two issues are likely to be addressed at least incrementally as the technology improves, although revolutionary improvements are unlikely. The read-only nature of a CD-R disc can be an advantage for applications such as data archiving.
The ubiquity of CD-ROM drives makes CD-R a useful means of transferring large amounts of data between systems, because anyone with a reasonably modern PC is likely to have a drive capable of reading the disc.
You can use a CD-R drive to produce relatively small runs of CD-ROM discs with large amounts of data that need to be transported to clients, customers, or branch offices. However, the CD-R media is relatively expensive (at $7 or so each), and recording a CD takes a considerable amount of time (several minutes to perhaps an hour, depending on your drive). These factors realistically limit data distribution via CD-R to perhaps a few dozen copies at a time. If you need larger numbers-say, 250 copies of your product catalog for distribution to customers-use a commercial CD duplication service.
Magneto-optical (MO) disks are another technology sometimes considered for use as a backup media. Magneto-optical disks use a combination of a high-power laser and a magnetic head to write to their media. The laser heats the media, allowing the magnetic head to realign the magnetic particles. Because this action is repeatable, MO disks are read-write like a traditional disk drive, rather than write-once like CD-R and WORM drives.
MO drives now have performance more similar to that of hard-disk drives than the performance usually associated with optical drives. However, MO drives have relatively low capacity and high media costs, making them inappropriate as backup devices for most situations.
Write-once, read-many (WORM) technology has been available longer than either CD-R or MO devices. WORM drives are available in various platter sizes up to 12 inches and in capacities of up to 6G per disc. WORM jukeboxes can provide near online storage capacity in the terabyte range. WORM drives are incrementally rewritable (data can be added incrementally), allowing backup of multiple versions of the same file or folder to a WORM disk.
WORM is an excellent-if expensive-archiving medium. For applications that require storing huge amounts of data, such as document imaging, the nearly online performance of WORM can be considered adequate for online use.
The Windows NT Server 4.0 backup application, NTBACKUP, has two obvious advantages: it's included with Windows NT 4.0 (so it's free), and, as a bundled application, compatibility and reliability problems are less likely to occur. Balanced against these advantages are NTBACKUP's paucity of high-end features and limited options. If your LAN is relatively small, the LAN's architecture is simple, and your backup requirements are modest, NTBACKUP suffices. For single-server environments that use a simple tape-rotation method and have no need to backup workstations from the server, NTBACKUP is more than adequate.
NTBACKUP can back up files stored on a drive using either the NTFS or FAT file systems, and it can restore the files backed up from a drive using one file system to a drive using the other file system. NTBACKUP does only file-by-file backups, and it makes no provision for doing a disk-image backup. NTBACKUP supports only tape drives as destination devices. You can't, for example, back up from one hard drive to another using NTBACKUP.
The NTBACKUP program files are installed when you install Windows NT Server itself. Before using NTBACKUP, however, you must first install support for your tape drive. Do so by following these steps:
8.13
The Devices page of the Tape Devices property sheet.
8.14
The Drivers page of the Tape Devices property sheet.
8.15
The Install Driver dialog listing manufacturers and tape device types.
Click the Have Disk button of the Install Driver dialog to install a new or updated driver provided by your tape drive's manufacturer.
8.16
The Files Needed dialog for installing tape backup drivers.
Using NTBACKUP to back up your files requires several steps. You must prepare and label the media; select the volumes, folders, and files to be included in the backup set; choose the appropriate backup options to use; and then run the backup itself. Optionally, you may choose to run a comparison pass after completing the backup to verify the integrity of your backup set.
Depending on the type of tape drive you use, there may be little-or quite a lot-of preparation needed. QIC tape drives require that tapes be formatted before use and that they be periodically retensioned to avoid breakage. The QIC formatting process takes considerable time, but such time consumption can be avoided by purchasing preformatted tapes. Like VCR tapes-the other technology that uses helical scanning-DAT and 8mm tapes require neither formatting before use nor periodic retensioning.
NTBACKUP gives you several tape tools, most of which you'll seldom need to use. NTBACKUP allows you to erase tapes by using either a standard erase (which simply deletes the header information) or a secure erase (which actually overwrites the data on the tape). NTBACKUP allows you to retension tapes that require such maintenance, and allows you to eject a tape if your tape drive supports software-controlled ejection. (Now there's a useful feature!)
If your labeling habits are poor, now is the time to fix them. The last thing you need is poorly labeled backup tapes. Inadequate tape labels make it much more likely that you'll eventually back up to the wrong tape. Also, poorly labeled backup tapes make it nearly impossible to find the correct tape days or weeks later when you need to do a restore. Make it a standard practice to label your tapes legibly and indelibly. Otherwise, you'll sooner or later regret not doing so.
How you label a tape depends on several factors, including the tape rotation method you use and the practices particular to your site. Make sure that the label is permanently affixed, and includes at least the server name, the volume ID (if appropriate), the set to which the tape belongs, and the tape number. For example, you might label a tape as follows:
- Admin Server
- SYS volume
- Set A
- Tape 1 of 1
to indicate that that tape is number 1 of 1 from Set A, used to back up the SYS volume of the Administration server. Alternatively, if you are of a minimalist bent, something like the following might do as well:
- \\ADMIN\SYS
- A-1
Always maintain a manual backup log, indicating the date that each backup was done, the contents of that backup, the type of backup, and the tape set to which the data was written. That way, if the server crashes, you won't find yourself trying to retrieve the software-generated backup log from a crashed hard disk drive.
Many people record this information on a label attached to the tape itself. A better practice is to keep a spiral-bound notebook nearby and record the details of each backup session in this notebook. This makes it easy to retrieve the information when needed without sorting through a pile of tapes.
After you prepare and label your tapes, the next step is to run Windows NT Backup and select the drives, folders, and files to be backed up. Proceed as follows:
Before trying to run a backup, make sure that you're logged on as either an Administrator or as a Backup Operator. Otherwise, you may not have the necessary permissions to access all files that need to be backed up, and the resulting backup tape will be incomplete.
8.17
The Backup - [Tapes] window, with a blank tape inserted in the tape drive.
If you have more than one tape drive installed, now is the time to verify that you have the correct tape drive selected. To do so, from the
Operations menu chooseHardware Setup. The Hardware Setup dialog appears, presenting you with a drop-down list of installed tape devices. Select the tape drive you want to use and click OK to select that drive.
8.18
The Backup - [Drives] window, displaying drives accessible to NTBACKUP.
8.19
NTBACKUP's folder view displaying folders and files available for backup.
8.20
Selecting specific folders and files for backup.
The check box for folders that aren't selected for backup are empty. The check box for folders in which all files and subfolders are selected for backup are marked with an [ts] on a white background. The check box for folders in which only some files and/or subfolders are selected for backup are marked with an [ts] on a gray background.
If you want to back up many, but not all, of the files on a particular drive, you can select those folders and files individually by marking their check boxes. An easier way, however, is simply to mark the check box for the drive itself, which appears at the extreme top left of the folder tree pane. Doing so marks all folders and files on the drive to be backed up. You can then clear the check boxes for those folders and files that you don't want to back up.
After you select the drives, folders, and files to be backed up, the next step is to choose backup options. To do so, click the Backup button in the Backup window to display the Backup Information dialog. The Backup Information dialog is divided into three sections (see fig. 8.21). The first section presents several items about the currently mounted tape, including its name, owner, and creation date. The first section also allows you to specify various options that determine how the backup job is run.
8.21
Setting backup options in the Backup Information dialog.
The following list describes the first set of backup options:
The second section of the Backup Information dialog, Backup Set Information, lets you specify information about particular backup sets, using a scroll bar to allow you to view one such set at a time. (The scroll bar appears only if you have selected more than one drive to be backed up.) Each backup set includes the following information:
The third section of the Backup Information dialog, Log Information, lets you specify logging information for the backup session, including the log file name and location, and the level of detail you want the log to contain:
After you select the options desired, click OK to begin the backup. The backup then begins, unless you have chosen the Replace option and the target tape already contains data. If so, you'll be prompted to make sure that you really want to overwrite the tape. When the backup completes, a compare pass begins, if you chose that option.
Regardless of the setting for Log Information, if NTBACKUP encounters corrupted files, it displays the problem in the status area of the Backup Status dialog and records this problem in the file CORRUPT.LST. If this happens, you first should determine the reason for it. After you do so to your satisfaction, delete CORRUPT.LST before trying to restore from that or any other tape. Failing to delete CORRUPT.LST results in error messages informing you of corrupt files each time you try to restore from that or any other tape.
As the backup begins, the Backup Status dialog appears (see fig. 8.22). The Backup Status dialog displays the progress of your backup until it completes.
8.22
Displaying backup progress and status messages in the Backup Status dialog.
The backup applet bundled with Windows NT Server 4.0 is a trimmed-down version of Arcada Backup Exec for Windows NT. If you like NTBACKUP but would prefer a product with more power and flexibility, consider purchasing Arcada Backup Exec. Arcada was recently acquired by Seagate Software Storage Management Group, and can be reached at http://www.arcada.com.
Unlike most third-party backup software, NTBACKUP has weak automation features. The biggest shortfalls are the absence of both a comprehensive macro language and the absence of built-in support for tape-rotation algorithms. NTBACKUP has no native scheduler, but it can be run from the command line by using the Windows NT scheduler service, AT commands, and batch files.
You can invoke all NTBACKUP options from the command line. This allows you to build an automated backup scheduling system of sorts, leaving you only with the problem of managing tape rotations manually. For a complete list of available commands and syntax, see the topic "Using Batch Files to Do Backups" in the NTBACKUP help file. If you need scheduling and other automation features, you're far better off buying a commercial backup software package instead of spending a lot of time trying to make NTBACKUP do things it really wasn't designed to do.
Restoring with NTBACKUP is relatively straightforward. NTBACKUP allows a backup set to be restored to the same system from which it was made, or to a different system. NTBACKUP also allows backup sets made from either of the supported file systems to be restored to a disk, using any of the three file systems. The following steps show you how to use NTBACKUP to restore files to your system:
8.23
The Backup - [Tapes] window, displaying the tape date and the folder backed up.
8.24
NTBACKUP displaying tape contents in folder tree format.
8.25
Selecting restore options in the Restore Information dialog.
The Restore Information dialog is divided into two sections. The Backup Set Information section displays properties of the mounted tape, including its name, the backup set of which it is a member, its creation date and its owner. The Backup Set Information section also allows you to set the following options:
The second section of the Restore Information dialog, Log Information, allows you to specify logging information for the restore session, including the log file name and location, and the level of detail you want the log to contain:
8.26
Windows Notepad displaying records created by a typical backup log.
After you select the options desired, click OK to begin the restore. The restore then begins and runs to completion, with a message that tells you the process is complete. When the restore completes, a compare pass begins (if you selected that option).
Choosing the right backup software is as important as choosing the right backup hardware. Windows NT Server 4.0's NTBACKUP is a quite competent application, but it's a bit Spartan. NTBACKUP provides fundamental backup features, but you won't find the bells and whistles on it like those provided by third-party backup software.
If you have multiple servers, plan to use Tower of Hanoi or other complex tape-rotation methods, or must back up workstations running diverse operating systems, you need to consider purchasing one of the commercial backup programs available from numerous third parties. Third-party software is needed if you must back up multiple servers in a heterogeneous network, such as a network that includes both Windows NT and Novell NetWare servers.
The rapid market growth for Windows NT Server 4.0 has backup software vendors scrambling to release versions of their products for Windows NT. Just a year or two ago, third-party backup choices for Windows NT were very limited, both in number and in feature sets. That situation has changed for the better and continues to improve with every passing month. Nearly all the big-name backup software publishers that previously concentrated their efforts exclusively on Novell NetWare have now released Windows NT Server versions; in some cases, the Windows NT version appears to have more features than the original version for NetWare.
Most of these software products, in addition to the complete selection of basic backup features you might expect, offer extensive customization options, powerful macro languages to automate the entire backup process, automated databases to support sophisticated tape-rotation algorithms, and the capability to back up workstations running diverse operating systems. Describing these third-party offerings is beyond the scope of this book and, because of the rapid pace of development of backup software, current products are likely to be upgraded for Windows NT Server 4.0. Vendor sites and magazine reviews on the Web are the best source of current information on third-party backup software.
Backing up critical data and assuring the ability to restore data when required are a network administrator's primary responsibility. Thus, this entire chapter is devoted to backup methodology, hardware, and software. If you're upgrading a network from an earlier version of Windows NT Server or from another network operating system, now is an ideal time to re-examine your data backup and restore practices, and to consider upgrading your backup hardware to greater capacity and faster data-transfer rates.
The following chapters contain material relating to data backup operations:
© 1996, QUE Corporation, an imprint of Macmillan Publishing USA, a Simon and Schuster Company.
