Disaster recovery

Because IT is the lifeblood of today.s e-business economy, it is
essential to be able to recover IT operations quickly after a
disaster. An IT disaster can be something as small as a failed disk
drive within a server or something as big as a data center burning to
the ground.

Other common examples of disasters include power failures, destruction
due to viruses, theft, sabotage, accidental commands issued by
administrators that result in files being deleted or modified,
earthquakes, insurrections, and so on. Being prepared for
eventualities such as these is the responsibility of IT management and
is critical to the success of modern businesses.

The key to successful recovery from these situations is a
comprehensive, tested disaster recovery plan. This is a business plan
that involves not just redundant hardware but personnel, procedures,
responsibilities, and issues of legal liability.

Implementation

The first step in such a plan involves risk assessment to determine
which components are most vulnerable and how the company can function
in response to loss of data or critical IT services. It is also
important to determine what is acceptable to management in terms of a
recovery window.many businesses today would suffer significant loss if
it should take more than a day or two to recover from failed systems.

Disaster recovery plans can be implemented in different ways depending
on the size and needs of the company involved. Three common approaches
to implementing disaster recovery plans are traditional tape backup,
e-vaulting, and mirroring.

The traditional approach of small to mid-sized businesses to disaster
recovery is to create full backups to tape daily and archive these
tapes off-site in secure locations. Most computer systems in
businesses use this method, primarily due to its low cost and its
wellknown procedures. However, tape backups must be verified, and test
restores must be performed periodically to ensure that this system is
in fact working. Furthermore, businesses must realize that it
typically takes up to 48 hours to restore a failed disk or system from
tape backup, a recovery window that may be unacceptable from an
e-business standpoint.

E-vaulting (or electronic vaulting or data vaulting) takes a different
approach. Instead of archiving data to tape, it is sent over
high-speed leased lines or Internet connections to a remote data
center for safe storage. E-vaulting makes it possible to do more
frequent backups and rapid restores, but the weak link is the wide
area network (WAN) connection, for if that goes down, a restore might
be difficult or impossible.

To make e-vaulting more effective, companies sometimes arrange for
backup servers to be running at the remote site and back up data
directly to those servers.  Then if a failure occurs with the
company.s main servers, control can be switched over to the backup
servers and business can continue without interruption. The problem is
that having duplicate systems in place is an expensive proposition and
complex to manage, and companies often try to save money by placing
older hardware at the remote site. When disaster strikes, this
hardware may not be able to perform as hoped and business will go
down.

Another version of e-vaulting involves using a mobile data center (a
network in a moving van) at the remote site. When disaster strikes the
primary site, the mobile data center can be brought on location and
managed by your staff.

Regardless of which form of e-vaulting is used, the main disadvantage
in the eyes of most companies is the cost, which is typically many
times that of using traditional tape backup solutions.

Extending the idea of e-vaulting is a method called mirroring, in
which identical hardware is placed at a remote site and data is kept
synchronized between servers in the primary and remote site. This is a
costly solution both from the point of view of the redundant hardware
required and the throughput needed for WAN links, but the restore
window is typically under an hour when disaster strikes, so it is a
good option from the perspective of e-businesses.  Financial
institutions such as banks and credit unions often employ this
solution to ensure maximum availability and reliability for customer
access to accounts.

Marketplace

While smaller companies tend to manage their own tape backup and
e-vaulting solutions in-house, large enterprises generally outsource
their disaster recovery needs, particularly if mainframe systems and
AS/400s are involved.  Three companies handle the lion.s share of
disaster recovery business at the enterprise level: Comdisco
Continuity Services, SunGard Recovery Services, and IBM.s Business
Continuity Recovery Services (BCRS).

E-vaulting companies abound and their range of services varies
greatly. One example of such a company is Imation, which offers its
LiveVault services for small and medium-sized companies.

Businesses that host their business services with Web hosting
companies often make use of disaster recovery services provided by
these services. An example is Exodus Communications, which provides
mirroring services for customers at remote data centers.  See Also:
backup, tape format



BACKUP

The process of making reliable copies of important data so that the
data can be recovered in the event of a disaster.

Overview

Performing regular backups is perhaps the system or network
administrator.s least glamorous but most important task. Data loss on
a corporate network can occur for various reasons, including

? Disk failures caused by hardware failure, power outages, or improper
use

? Network problems leading to lost packets that are not acknowledged
because of router congestion or other situations

? Virus infection, resulting in corrupted files

? Sabotage by hackers or disgruntled employees, resulting in erased
data

? Theft of hardware from the premises

In each of these scenarios, having reliable backups of your company
data is essential to recover from the disaster and continue normal
business functioning.

At the enterprise level, backups can be performed using a variety of
technologies, each of which have their own advantages. These
technologies are a blend of backup device hardware and how these
devices are implemented.

The next section of this article looks at a few common
scenarios. First, backup solutions can be characterized by the devices
used to store the backed-up data. These devices can include

? Tape drives: A tape drive is a device that stores data on magnetic
tape. Many kinds of tape drives and tape formats are supported by
different vendors, and these are discussed more fully in the articles
.tape drive. and .tape format. elsewhere in this book. Generally
speaking, however, tape drives have capacities in the tens of
gigabytes range and are suitable for backing up data from individual
servers or small groups of servers.

? Tape libraries: A tape library consists of a set of tape drives, a
large collection of tapes, and a robotic mechanism for loading and
unloading tapes into drives. Tape libraries are common in large
companies and can typically store several terabytes (one terabyte
equals 1000 gigabytes) of data from groups of servers. For more
information, see the article .tape library. elsewhere in this book.

? Optical drives: Another medium for backing up data is optical
drives, which range from simple CD-R/W drives to DVD-W drives and
libraries containing many such drives. Optical drives are not as
common as tape drives and libraries.

? Storage appliances: These are generally rack-mountable black-box
solutions in which the underlying operation is not important. Storage
appliances are generally used for live data storage but can also be
used for small-scale backup purposes.

? Storage Area Networks (SANs): While SANs are primarily used for live
storage of data, they can also be used for archiving backup data. See
the article .storage area network (SAN). elsewhere in this book for
more information.

Besides these different backup devices, there are also various ways of
implementing them for backing up data from network servers:

? Server-based backups: In this scenario each server that holds
valuable data has a tape drive directly attached to it, usually
through a Small Computer System Interface (SCSI) connection. The
disadvantage of this scenario is that it scales poorly for large
companies.administrators would need to run around each morning to
collect tapes from drives scattered all over the network.

? Network backups: This is the most common scenario in most large
companies. In a typical network backup scenario, a group of servers on
a local area network (LAN) are connected using a second network
interface card (NIC) in each server to a separate LAN dedicated for
backup purposes. This dedicated backup LAN is concentrated using a
Fast or Gigabit Ethernet Switch, which is also connected to a
dedicated server called the backup server. The backup server has
special software running on it that initiates and manages the job of
backing up data on the production servers.

The backup server itself is then connected by SCSI or FiberChannel to
a tape library (see illustration).

? LAN-free backup: This is a simplified form of network backup in
which there.s no second backup LAN. Instead, fiber channel cards are
added to servers needing to be backed up and these are connected using
fiber-optic cabling to a fiber-channel router, which then forwards the
information directly over Fibre Channel links to tape libraries (see
illustration).  LAN-free backup is an emerging approach that is
gaining in popularity due to its simplicity and high performance.

? Serverless backups: This is a further refinement of LAN-free backups
that takes the actual task of processing the backup from the servers
and moves it to a Fibre Channel switch or router used to connect the
servers to the tape libraries. This can provide significant relief to
the servers since generating backups is a processor-intensive and
memory-intensive job that limits other functions they can perform
while the backup is occurring. Serverless backup solutions are just
emerging in the marketplace.

? Storage over IP: This technology backs up data from network servers
directly to backup devices such as tape libraries and SANs using only
an Ethernet network. No backup server is required to convert the data
from Ethernet frames for transmission over SCSI or Fibre Channel
connections to the backup device. Storage over IP is an emerging
technology that promises to have a large impact on the backup market,
and it is discussed further in the article .storage over IP. elsewhere
in this book.

? Internet backups: Backups can also be outsourced over the Internet
to a Storage Service Provider (SSP) that is responsible for managing
actual backup hardware and securely storing your data.  For more
information on this, see the article .televaulting. elsewhere in this
book.

Finally, a third component of a backup system is the backup software
itself. Some of the more popular backup software products used in the
enterprise include
? ArcServeIT from Computer Associates
? Backup Exec and NetBackup from VERITAS Software Corporation
? Legato NetWorker from Legato Systems
? Storage Manager from Tivoli Systems
? Backup Express from Syncsort
? Hiback and Hibars from Hicomp Software Systems

Implementation

Instituting a regular backup plan is one of the main components of a
company.s disaster recovery policy (see the article .disaster
recovery. elsewhere in this book for more information), and the
importance of doing so cannot be stressed enough. To guard against
these unexpected losses of data.or rather, to prepare for them, since
they are, to a certain extent, inevitable.establish a disaster
recovery policy that includes a reliable backup plan. In today.s
business world, where data is the lifeblood of the enterprise, a
comprehensive plan is essential. The following steps are recommended
when creating such a plan:

? Decide what kind of backup storage devices to use.  Options range
from small digital audio tape (DAT) drive units capable of backing up
several gigabytes of data to large automated tape libraries capable of
handling terabytes of centralized data storage.
Other backup options include optical storage libraries and removable
disks such as Iomega.s Zip drive disks or Imation SuperDisk disks.

? Decide whether to back up servers with dedicated, locally connected
storage devices or over the network to centralized backup
libraries. Network backup systems suffer from a single point of
failure (the network itself) but are simpler to administer than a
multitude of individual backup units.

? Decide whether individual users. workstations should also be backed
up. A more cost-effective option is to educate users to always save
their work on a network share located on a server that is regularly
backed up.

? Decide how to secure the storage of backup tapes and other
media. Will duplicate copies be stored both on-site (for easy access
if a restore is needed) and off-site (in case the company.s building
burns down)? Make sure the storage facilities are climate-controlled
and secure.

? Decide what kind of backup strategy to employ. A backup strategy is
a combination of a backup schedule and various backup types, including
normal, copy, incremental, differential, and daily copy backup
types. Also consider whether you will verify all tapes immediately
after each backup is performed. For further information, see the
articles .backup strategy.  and .backup type. elsewhere in this
chapter.

? Assign various aspects of the backup procedure to the responsible
party. One option some companies now use is to back up data over the
Internet to a third-party backup service provider that stores and
maintains the backed-up data. This method involves issues of trust and
of the Internet connection as a point of failure.

? Test backups periodically to ensure that they are actually
readable. Nothing is worse than thinking you have a backup when in
fact it is unreadable.

Notes

To enable administrators to perform regular backups, Microsoft
Corporation includes backup utilities with all