Protecting Against All Possible Disaster Events
Before evaluating data protection products, carefully consider the many types of disasters that threaten your organization. For most companies, the optimal disaster recovery program incorporates multiple technologies that work together to protect all business-critical information from the full range of events that can cause data loss.
One useful way to categorize potential disasters is by the reach of their impact across the enterprise. A single disaster event can cause data loss within a system, building, site or geographic region. The following chart plots numerous disaster events by the reach of their impact. For example, a fire is shown to affect a system, a building and a site, but not an entire region.
For each possible disaster, consider the frequency that your organization will be exposed (monthly, annually, every five years?) and the business impact of the event. Responding to a large-scale disaster that disables critical business services to many users is typically much more important than correcting a single-user hardware failure.
Selecting Data Protection
Three technologies currently are available for disaster recovery. Tape backup and fault-tolerant storage devices are well-known and utilized by thousands of organizations worldwide. These mainstay products provide a means to recover from disasters that affect an individual system or building. Network data mirroring is a newer technology, complementary to tape backup and RAID devices, that affords protection against many disaster events that can affect an entire site or region. Network data mirroring compensates for the deficiencies of tape backup and fault-tolerant storage devices, and rounds out a complete solution for organizations requiring robust data protection capabilities as part of their disaster recovery program.
Network Data Mirroring
Network data mirroring duplicates the full or partial contents of a disk from one system to a disk on another system by sending data over a local- or wide-area network as the data changes. Comprehensive network data mirroring products, such as Qualix DataStar and Octopus DataStar from the Qualix Group, mirror user files, application files and database information for UNIX and Windows NT servers, respectively. Updates to all databases and files are copied as they are output to the source system disk, regardless of whether or not the files are open and in use on client machines. By continuously mirroring data to target systems at remote sites, network data mirroring accomplishes three important objectives.
Objective One: Protect Against Site- and Region-wide Disasters
Disasters that affect entire sites or regions - including earthquakes, floods, hurricanes or extended power outages - can render traditional backup technologies useless, even if fault-tolerant devices are employed. Any disaster event that reaches beyond the single building level threatens to destroy not only the primary, operational system, but also the redundant hardware and its backup data. By maintaining a copy of mission-critical data on a target server geographically removed from the source server, network data mirroring eliminates the risk that a site- or region-wide disaster will damage both original and backup data and systems.
Reduce or Eliminate Data Loss
With network data mirroring, critical business information is protected continually, with little or no latency between the source and target systems. This is crucial for organizations, such as financial institutions, that require recovery of complete, up-to-date business information following a disaster. Synchronous mirroring (see sidebar), while degrading application performance somewhat, maintains exact duplicates of all source files at all times. Asynchronous mirroring allows for a small amount of latency between data on source and target systems for the sake of improved application performance. Either configuration reduces data loss in the event of a disaster.
Recover Data Quickly
Following a Disaster
Since network data is mirrored to an active system, critical information can be recovered rapidly without restoring the data from an off-line source. The easy recovery of the mirrored data is particularly important to companies that need to restore operations within mandatory windows of time, sometimes as short as minutes or hours. While some degree of performance degradation may result from shifting core business operations to secondary systems, most organizations find this a small price to pay to remain operational in the face of a wide-area disaster. The target system can be used as a functional replacement until primary facilities and systems are ready to resume operations.
Network Data Mirroring
Augments Tape Backup
Tape backup, the most common form of data protection, effectively protects against inadvertent file deletions and provides some degree of protection against hardware failures. If tapes are securely stored off-site, data should be protected from any type of disaster event that can damage local storage devices. Tape backup is inexpensive and simple to operate given recent advances in tape capacity and automated tape libraries.
While most organizations benefit from tape backup, the technology does not meet all requirements for robust disaster recovery. Most importantly, data on backup tapes is not current and organizations risk losing all data created or modified since the last backup. Even for companies diligently performing daily tape backups, the cost of losing information processed during the business day can be exceedingly high. In addition, if tapes are utilized on-site to capture system updates, business data is subject to loss from events that damage both the primary storage system and the tape backup media itself.
Network data mirroring provides disaster recovery capabilities superior to those of tape backup. Restoring data from backup tapes to primary system servers can require hours or even days, depending on the amount of data, the tape device employed and the size of the connection between tape drive and server. Compounding this delay is the fact that the most recent tape may not be available at the disaster recovery location, in cases where companies are able to operate a recovery location distant from the primary location. In addition, the human labor required to load tapes and administer data backup on remote systems can be costly compared to automated methods such as network data mirroring. Network data mirroring reduces the time required for organizations to recover data from disasters of all types.
Network Data Mirroring
Redundant Array of Independent Disk (RAID) devices group several disks together to act as one large disk drive. This technology was designed to deliver improved disk performance and fault tolerance by continuous disk mirroring. RAID level 1, the simplest configuration providing data protection capabilities, consists of a primary disk and a redundant physical disk drive, typically co-located with the primary disk. If the primary disk fails, the redundant disk takes over, transparent to the computer system and applications.
However, RAID arrays only protect corporate data against isolated hardware failures. RAID devices eliminate latency of the backup data, but they still fail to protect organizations from disaster events that affect entire buildings, sites or regions.
In most fault-tolerant storage configurations, primary and redundant disks are housed in the same location. Any event that damages both devices will cause the organization to lose both primary and backup data. Disaster events of this type are numerous, including building-level events such as burst water pipes, site-level events such as fires, and region-wide events including hurricanes and earthquakes. These far-reaching disasters jeopardize business-critical data even if RAID devices are in place. Network data mirroring provides complete protection from these types of disasters because data is copied to an active system at a remote site.
Implementing the Optimal
Disaster Protection Program
Small businesses often can achieve adequate data protection with tape backup alone. For most organizations with significant amounts of business-critical data, however, the right disaster recovery program incorporates several different technologies. Tape backup is almost always mandatory because the costs are low and the technology provides a chronological file archive and protection from inadvertent file deletions. More advanced mirroring technologies do not offer this fundamental protection since a file deletion on the source system will be automatically duplicated on the target system.
Any organization that requires high availability and no data loss from device failures should, in addition to tape backup, employ RAID disk mirroring technology. RAID devices provide fast disk access and protection against storage device failures, which are among the most common types of failures. RAID eliminates backup data latency and provides fault-tolerant storage that both protects data and provides high availability with minimal human intervention. Tape backup and RAID devices, however, leave organizations at significant risk of data loss in the event of a wide-area disaster.
For companies that require protection from the full range of possible disaster events, network data mirroring provides the most comprehensive data protection when combined with tape backup and RAID devices. To implement the optimal disaster recovery program, answer the following questions before evaluating the wide range of data protection products:
What are all the potential disasters that threaten your organization?
What is the relative risk of each disaster event?
How valuable is your business data?
How much critical data does your organization have?
How frequently does that data change?
Evaluate data protection technologies in terms of how well they individually and collectively meet your organization's requirements for both data integrity and recovery speed. The key to success lies in deploying the optimal mix of technologies to protect your organization from all possible disasters that threaten your business-critical information.
Network Data Mirroring Overview: What it is
Network data mirroring is a newer technology for data protection and disaster recovery applications. Network data mirroring compensates for many of the deficiencies in tape backup and RAID devices, and rounds out a complete solution for organizations requiring robust disaster recovery capabilities.
Network data mirroring technology accomplishes three important objectives:
Protect against site- and region-wide disasters
Reduce or eliminate data loss
Recover quickly following a disaster.
How it works
Network data mirroring replicates the write operations to a disk on one system to a disk on another system by sending the changes over a local-or wide-area network.
Two types of mirroring architectures exist. With synchronous mirroring, any write operation initiated on the source system is not completed until the corresponding operation is completed on all designated target systems. This 'guaranteed mirroring', while ensuring an exact copy of all data at all times, degrades application response time and introduces the network and target system as single points of failure. This architecture is generally only recommended in cases where preventing the loss of even a single transaction justifies the tradeoffs of slower applications, and where corporate networks are extremely reliable.
With asynchronous mirroring, target systems are updated immediately after an operation is completed on the source system. Since the source system I/O does not need to wait for write operations to complete at the target, there is minimal impact on application response time. Despite the fact that there is always some lag time between source and target data, asynchronous mirroring can approach the zero latency of synchronous mirroring if properly implemented.
Jason Welco, senior product line manager at the Qualix Group (San Mateo, CA), has helped pioneer the development of network data mirroring technology. He has more than 16 years of experience with leading information technology companies, including Siemens, IBM and Rolm. He may be reached at email@example.com