Disk-to-disk replication can offload the backup process from production servers.
By Dave Demlow
IT professionals who earned their seat at the corporate strategy table during the Y2K run-up now should be helping business-unit peers avoid any post-Y2K complacency about a looming dilemma involving data backup. Corporate executives with responsibility to protect company assets such as data must be knowledgeable about the continuous- replication option for updating readiness and accommodating increased access demand and more-robust server capabilities.
The reality is that tape backup systems, which have been sufficient for generations, have not kept pace as 24x7 demand for data access has become universal and the burden of backing up higher-capacity servers has multiplied. Continuous-replication technologies that enhance tape backup now offer suitable solutions.
Continuously replicating critical data to centralized servers means tape backup systems can back up the disk-based replicas, rather than making administrators back up from production servers-a system-clogging, bandwidth-hogging process. Meanwhile, tape backup retains its priority function as an archival medium, maintaining its investment value.
This article discusses how disk-to-disk replication technologies can provide an efficient answer to the business-continuance challenge and considers why complacency has left many companies vulnerable despite warnings.
Narrowing backup windows
Since the advent of the mainframe era, it was acceptable to bring servers to their knees each night in a "race to sunrise" to duplicate data on tape backup before users returned to their desks. Multiple copies of data were made after it was determined how many were needed, where to put them, and how often to update them. The tapes generally were sent to off-site storage sites to be retrieved only if an unexpected outage called for the data to be reconstructed.
Obviously, this left at least a day's worth of data unprotected from a server or disk crash, and it took several hours to days to rebuild the server from tape. Gaps in access and recovery were even wider with weekly backups (or combinations of weekly full backups and daily differential backups). For many companies, this was acceptable. But increased data loads on corporate networks now are making it competitively unacceptable.
Continuous disk-to-disk backup can be augmented with traditional tape-based backup.
That backup window keeps closing. Accelerated by Internet-based communications, normal business processes are evolving quickly. Distributed employees, partners, and customers are logging into servers and databases globally 24x7 while e-commerce enterprises need always-on systems and continuous data protection for order processing. Even e-mail is now generally considered "mission-critical," with delays able to severely impede a company's ability to operate.
But despite increases in megabytes-per-minute speed, tape-backup technology has not kept up with the growth in server size so that total restoration times continue to increase. A typical 50GB server, for example, could be restored in about eight hours at a rate of approximately 10MB per minute with traditional tape-backup technology, if all worked well. Even with the best tape backup available today, restoration for a 1TB server at a 900MB-per-minute, or 52GB-per-hour, rate will take 19 hours. During that time, of course, the server and application would be unavailable to users and loom as a threat to service level agreement (SLA) delivery.
Using disk-based, continuous-replication technology to enhance tape backup, however, can accommodate the realities of swelling disk capacity and the shrinking backup window. Having a live replica on disk avoids having to restore.
Opening the backup window
The backup window can be opened as wide as needed using disk-based continuous-replication technologies. The production server is unaffected because backup tapes are made of the disk-based replicas. This shifts periodic tape backup from multiple production servers to a dedicated backup server, making possible truly centralized tape backup.
The process of continuous replication can be at either the file level (replication) or disk block level (mirroring). File-based replication, unlike traditional tape backup, uses the smallest possible amount of bandwidth by capturing and transmitting only changes, rather than sending the whole file across the network each time. Disk mirroring, by comparison, maintains a replica of databases and/or file systems by capturing and transmitting changes at an entire disk block level. Its usually synchronous nature also delays the I/O process of the primary server to handle the update of the remote site.
Asynchronous replication has the advantage of significantly reducing the need for network bandwidth, particularly for large files like databases, video captures, and activity logs. Unlike disk-mirroring applications, file-replication software is flexible. Users can select only the files or directories they deem mission-critical and replicate those continuously. Because mission-critical files typically account for just a fraction of an organization's total data, users can significantly reduce their off-site data-protection costs. This significantly reduces the cost of off-site protection, compared with disk-mirroring applications where it's "all or nothing."
Continuous replication allows a consistent copy of an application environment to be accessed by software utilities (or other applications) while the production copy stays online. A dedicated high-availability (or disaster-recovery) server can be accessed even if the production server goes down. Regardless of its state on the source server, each file on the target server is closed and available for backup at any time. Changes stay queued on the target until backup is done and then are applied automatically. This means that backing up the secondary server has no impact on the production server's performance and generates no additional network traffic.
Tape backup remains the most popular method of protecting enterprise data from unplanned outages. But disk-based replication solutions are gaining as their application expands beyond disaster recovery to operational functions, including meeting SLAs. Gartner Inc. forecasts that by 2003, 75% of large enterprises are expected to be combining disk-based data replication and tape-based technology for rapid application recovery.
Disk-based continuous-replication tech nologies present a scalable and easily managed option for enhancing tape backup systems. They can provide "instant" protection along with fast recovery via online disk.
The role of tape backup then becomes maintaining archival, point-in-time copies of different file versions for the record. Continuous replication also can be leveraged to serve a multitude of business needs, including high availability and fail-over capability, disaster recovery, Web-content replication, data distribution among multiple servers, platform migration, and change management and testing.
Some CIOs responsible for protecting company data, however, may be over- estimating the ability of tape-based systems to keep up with both user-access demand and server growth. Surveys show that complacency in this regard is becoming more widespread. The celebrated Y2K compliance challenge brought money and attention to upgrading business-continuity and data-recovery capabilities, but apathy over the need to keep pace now threatens to adversely affect productivity, competitiveness, customer relations, and revenue.
A Gartner study showed that most IT managers publicly support the criticality of business continuance but said apathy crept in when budgets started tightening and there were no recent catastrophic failures. Meanwhile, turnover of IT personnel often pushed business continuance to a second-rate problem for "my successor," which echoes arguments in the mid-to-late 1990s for delaying Y2K renovations. "If they don't feel the pain," one Gartner symposium exhibitor said, "they won't buy the cure."
"Like any form of insurance, the decision ultimately boils down to risk evaluation and how much is acceptable," the Gartner report said.
"Cost is no longer a viable excuse, as there are many products that focus on the vulnerable server-based environments that are amazingly cost-effective to acquire and deploy. Considering that most of the outages occur with these distributed platforms, protecting your enterprise has never been less expensive."
David J. Demlow is vice president of product management at NSI Software (www.nsisoftware.com) in Hoboken, NJ.