|
XactCopy White Paper
Summary |
| The purpose of this white paper is
to provide an overview of a new technology approach toward
fast system recovery for servers and mission critical
workstations running the Windows NT Operating System (OS). The
paper illustrates a simple and inexpensive way of recovering
from OS and hard disk failures in minutes instead of hours.
Implementation requires a second hard disk and DuoCor's
XactCopy™ software.
|
Overview of Backup Methods |
| Most full-system backup products
take hours to restore a failed system to normal operation. In
many environments, downtime is intolerable, yet striking a
balance between backup time and restore time is an issue that
is unique to each environment. All of the backup and restore
options given below are analyzed to indicate how the new
DuoCor technology is most suitable where system downtime, due
to either backing up data or restoring it, is intolerable.
|
There are three different types of
backups: full backup and two types of partial backup called
incremental and differential.
- Full Backup - A full backup usually includes all of
the system and data files contained on the system drive. The
best form of full backup is a sector-by-sector copy to the
target storage device because the single copy provides the
fastest system recovery. Most disaster recovery plans
recommend performing a full backup at least weekly.
- Incremental Backup - With incremental backup, the
operation includes only those files changed since the last
full or incremental backup. Incremental backups take less
time to perform because of the reduced amount of data being
written to the target storage device. A full system recovery
takes longer to accomplish because the process begins with
the last (most current) full backup followed by all
subsequent incremental backups.
- Differential Backup - With differential backup,
every file that has changed since the last full backup is
backed up each time. Compared to an incremental backup, it
is much faster to restore from a differential backup because
the last full backup and the last differential backup are
the only copies necessary for the task.
With any of
these three backup types, either individual file or disk image
methods may be used for the backup process:
- File-by-File Method - The file-by-file method
requests each individual file and writes it to the backup
device. For full-system backups, the backup time is much
longer using the file-by-file method over a sector-by-sector
disk image method.
- Disk Image Method - The disk image method is a
sector-by-sector identical copy of the entire system disk.
The image backup process typically does not care what is on
the system disk or even what it is doing at the time of
backup. Disk imaging is much faster than the file-by-file
method. If an operating system or disk failure occurs,
restoring the system from the duplicate image medium (often
a tape cartridge) offers the fastest method of
recovery..
The primary reason why system administrators perform
full-system backups is for their use in recovering a system
after operating system failure, hard disk failure, or
significant data loss. Full-system backups differ from
archive, which is the method of long-term or legally required
storage of certain data files from day-to-day.
Tape backup systems are the predominant choice for the
various backup methods. Although tape backup offers the best
solution for archiving files on a periodic basis, its use for
full-system backups is less desirable because the time to
recover from a failed system is relatively long.
|
Overview of DuoCor Technology |
| XactCopy's primary function
addresses full-system backups for the purpose of immediate
system recovery. There are two important differences
between XactCopy™ and other full-system backup methods:
- XactCopy's routine full-backups are very fast (typically
under 3-minutes), which promotes more frequent use.
- XactCopy™ utilizes a dedicated disk drive as the backup
medium, which offers instant recovery from OS or drive
failures-the backup drive is bootable directly.
XactCopy™ makes an identical
sector-by-sector copy of the system drive to the backup drive.
In the XactCopy™ program and in this paper, we refer to the
dedicated secondary drive as the Data Protection/System
Recovery (DPSR) drive. The DPSR drive remains invisible to
the operating system at all times, rendering data safe
from alteration or corruption.
Following a system drive or OS failure, the DPSR drive is
booted directly without having to rely upon floppies, partial
OS restore, slow full-system restore from tape (after disk
drive replacement or repair), or complicated and time
consuming incremental tape restores. XactCopy™ places the
system back into operation almost immediately, which enhances
productivity with system up time.
After the initial sector-by-sector backup, which occurs
during program installation, subsequent (routine) full backups
are similar to an incremental backup. Only those files changed
since the last backup are a part of the periodic update. The
ability to use incremental updates, which enhances the speed
of the backup, is unique to the choice of backup medium used.
Because the backup device is similar to the hard drive that it
is protecting, it is possible to compare data between the
drives to search for all changes made since the last full
backup. This incremental disk backup results in a full backup
of the system drive to the DPSR drive.
XactCopy™ DPSR is a fast alternative method to performing
full system backups without tape, for immediate system
recovery when needed. It is not a replacement for incremental
tape backups, which companies generally use for legal and
other reasons.
All backup operations with XactCopy™ occur from within the
operating system, which means that the server or workstation
remains live. Most all other disk-to-disk-based copying
programs require the administrator to shutdown the server or
workstation and boot from a DOS prompt to run them.
The steps listed below illustrate a full system recovery
following an operating system or drive failure:
- Remove the failed or non-bootable System disk, or
change the boot sequence as applicable to the
installation.
- Reboot the system from the secondary DPSR drive.
|
Applications of the Technology |
| One of the most frequent questions
about XactCopy™ is its application with hardware mirroring, NT
mirroring, and RAID. An important distinction about mirroring
and RAID is that deleted or corrupted files on the system
drive concurrently write to the secondary drive or drive
array. Random disk arrays and mirroring only protect against
drive failure: they do not protect against file problems.
When a critical system file becomes corrupted, such as with
the NT "Blue Screen of Death," the disk array offers no
benefit for system recovery. Typically, operating system
failures occur more frequently than drive failures and
protection from OS failures with XactCopy™ is possible because
the user decides when to write to the DPSR drive. Even when a
routine backup is scheduled, the backup cannot occur if the
operating system has failed.
|
NT Servers with RAID |
| In this configuration, XactCopy™
provides almost instant recovery of the NT server following
non-recoverable operating system failure. If the boot
partition is located on the RAID, the application entails
transferring it from the RAID onto a separate small SCSI or
IDE drive. After successfully moving the boot partition, a
second installed drive becomes the DPSR drive, which protects
the new system (boot) drive. XactCopy™ is used to transfer the
primary boot partition from the RAID to the new system boot
drive and also to copy its contents to the secondary drive on
a periodic basis determined by the system administrator.
After accomplishing the reconfiguration, XactCopy™ performs
periodic copies of the entire contents of the primary boot
drive without booting from DOS, which means that the server
continues to operate. All routine backups are incremental
(changed files-only) and result in a full backup to the
secondary small DPSR drive.
When the server encounters a non-recoverable operating
system failure, the system administrator can immediately boot
the backup drive to restore system operation. Total downtime
is typically less than a few minutes and because of its
simplicity, a non-skilled technician can handle the recovery.
If the primary drive is housed in a removable bay, the
recovery procedure is to physically remove the primary drive.
If the primary drive is not in a removable canister, changing
the boot address recovers the system. Figure 1 illustrates
adapting the configuration for optimal OS failure recovery in
a RAID environment.
|
NT Servers with Mirroring |
| There are two basic types of
mirroring: hardware mirroring (with an installed special
hardware card) and the software mirroring available in the
Server version of Windows NT, and other third party vendors.
If a system configuration is set up under NT or another brand
of software mirroring, discontinue using the second drive
under the software or hardware mirroring scheme and substitute
this drive as the DPSR drive with XactCopy™.
With XactCopy™ installed, system recovery is possible from
both types of failure-disk drive and operating system
problems-where the latter was not previously available. An
additional benefit from this configuration is that of gaining
protection from non-system file corruption, deletions, and
possibly virus infections.
|
NT Servers without RAID or Mirroring |
| In this application, the DuoCor
technology provides fast recovery from both OS and hard drive
failures. Periodic full backups of only those files changed
since the last backup, take place from within the operating
system in approximately one to three minutes-while the system
is running. The system administrator has the option to
perform periodic full backups automatically by using the
XactCopy™ Scheduler Service (an NT Service) or manually at any
time.
The technology offers a low cost alternative to RAID for
drive failure protection plus the addition of OS failure
protection. Frequent updates of the system drive ensures
up-to-date DPSR drive data, which minimizes data loss and
enhances fast system recovery. This configuration also
protects from corruption and loss of data files, which are
other than critical system files.
XactCopy™ also restores files, folders, and complete
partitions very quickly. The main screen of the program
displays the contents of both drives in a side-by-side
Explorer-like fashion. To aid in quickly identifying file
differences between the System and DPSR drives, the program
places a red colored not-equal sign next to the file. Files
deleted since the last backup appear in the DPSR panel and not
in the System drive panel.
By highlighting the file or folder and clicking the
Restore Files button, the program instantly
restores the file or the entire contents of a selected folder.
Using the full-partition restore command of the program
quickly restores an entire partition.
|
Mission Critical Workstations and Stand-alone PCs |
| The application of XactCopy™ at a
mission critical workstation is identical to that of its
application on a non-RAID or mirrored server system. The
technology offers protection from loss of mission critical
data and its immediate recovery without the need to search
through a tape library or network server. Like its server
counterpart, the program offers fast system recovery from OS
or system drive failures.
In many instances, backing up data at the workstation level
has the added benefit of reducing network traffic. Another
advantage afforded by the fast system recovery features of
XactCopy™, is that of productivity for the workstation user.
With different schemes for servicing a failed workstation,
which range from replacement to complete rebuilding,
XactCopy's instant recovery feature does not noticeably
interrupt the workflow of the user. The administrator or third
party service organization can delay repair of the system to
off-hours or when time permits.
|
Backing-up Open Database Files |
| When performing backup operations
with XactCopy™, it copies all open database files on a
sector-by-sector basis. With several workstation users
changing information and using a sector copy technique, the
database would normally be uncoordinated resulting in a "dirty
backup." To solve the problem of "dirty backups," the server
version of XactCopy NT contains an Open Transaction Manager
(OTM™), which provides a "clean backup" of all open files
while users are changing information on the open files.
How XactCopy™ and OTM™ work Together
OTM™ presents a stable, non-changeable picture-in-time of
any system hard drive to the DPSR drive by creating an
alternate "virtual drive," or static copy of the drive to be
backed up. When OTM™ is started by XactCopy™, it waits for a
short period of inactivity (5 seconds) where no writes are
occurring to any of the volumes or drives that have been
selected for backup. Once this quiescent period is obtained,
OTM™ is enabled and maps-in a virtual drive letter for each
volume selected to be backed up. XactCopy™ accesses this
static virtual volume, instead of the original volume, which
is changing during the backup.
When a write command occurs on the original volume, OTM™
pauses it and copies the old corresponding data to its cache
file and immediately sends the original write data to the
system drive. This action keeps the system drive current and
unaffected at all times during the backup. Read requests from
all applications except the backup are passed directly to the
system drive with no intervention. Read requests from
XactCopy™ are passed to the OTM™ filter driver, which
determines if the requested data is already in cache. If data
is in cache, OTM™ passes the cached data to the DPSR drive. If
not, the data is passed directly from the system drive. Since
OTM™ only needs to preserve the original data, additional
writes to the same sector are not cached and are passed
directly to the system drive. (For additional information and
details, see the OTM™ White Paper on DuoCor's Website.)
|
The Benefits of Increased Backup Frequency |
| By performing frequent backups of
the database or other applications, data is kept more current
resulting in less data lost in the event of a catastrophic
failure. In any backup environment, a need to restore
non-current data after a critical failure exists because of
the difference in time between the failure and the last
backup. The data loss equation is:
Data Loss = Time of Failure - Last Backup
To minimize data loss, the system partition should be
physically separated from the data partition(s). The system
partition should only be backed up when new applications are
installed, new users are added, or any other changes that
affect the operating system's registry. Other than for the
purpose of duplicating these registry-type changes, frequent
backups of the operating system partition are not needed.
Protecting data partitions through frequent backups is
another matter. According to the data loss equation, frequent
backups of the data partition(s) results in less information
lost after a critical failure occurs. XactCopy™ allows
partition selection for manual or automatic backups to
accommodate this scheme for minimizing data loss.
|
An Alternate Scheme for Zero Data Loss when OS Failures Occur |
| Systems configured with the
operating system and data partitions on the same physical disk
drive, as discussed in the previous section, still remain
vulnerable to data loss. Suppose that 45-minutes after a
backup operation, the operating system fails and becomes
non-recoverable. Booting the DPSR drive will recover the
system, but the last 45-minutes of data will not be on the
DPSR drive. These data can be copied from the drive where the
operating system failed, but the process will consume time to
accomplish.
By separating the system partition from the main data drive
(as in the RAID application discussed above) and placing it on
a separate small IDE or SCSI drive, system recovery issues
become separate from rapidly changing data activity on the
main data drive. Figure 3 illustrates a typical configuration
maximized for fast system recovery and zero data loss.
For maximum data protection and minimum downtime, the
scheme requires the addition of three hard drives:
-
One small IDE or SCSI drive to house the primary boot
partition.
-
A second small IDE or SCSI DPSR drive to protect the
primary boot (System) drive.
-
A third DPSR drive to protect the main data drive.
Unless changes occur to the Windows Registry, there is no
need to perform frequent backups of the operating system
(boot) partition. If the operating system becomes
non-bootable, boot the DPSR drive for immediate system
recovery. Frequent backups of the main data drive will protect
against data loss resulting from a drive failure, corrupted,
or deleted files and folders. This increase in backup
frequency results in more up-to-date data and a corresponding
lesser amount of data needed for recovery from one's
incremental tape backups.
|
Technology Benefits Summary |
- In the network server, workstation, and stand-alone PC
environment, the technology does not require shutting down
the system to run manual full backup copies of everything on
the system drive. The result is more frequent use and
less data potentially lost.
- Offers almost instant system recovery; reboot from the
DPSR drive without the use of DOS utilities, OS reloads,
floppies, or complicated and time consuming incremental tape
restores. The result is increased productivity and the
associated cost of system downtime.
- Provides a low cost alternative to RAID servers for
protection against both system disk and operating system
failures. This results in cost savings and increased
disaster recovery protection.
- "Hidden" secondary disk cannot be altered by the user
or corrupted (or changed) by the operating system. No drive
letter conflicts to worry about.
- Protects against lost or corrupted files by allowing for
immediate restoration of files, folders or full partitions.
The result is increased productivity.
- Significantly reduces or eliminates data loss
trauma and its associated affect on business efficiency.
|
Conclusion |
| With the ever-decreasing cost of
disk drives combined with the ever-increasing cost of downtime
and reconstruction of lost data, the DuoCor DPSR
Immediate Disaster
Recovery technology has a place in most
enterprise systems.
|
|
