Tape pipeline for BC/DR planning: new considerations in backup and recovery. (Storage Networking).

Tape has long been an affordable backup medium. However, when it came to remote data replication, tape had severe limitations. With heavy latency impact over as little as a 10-mile distances, remote tape backup was not a viable option for business-continuity/disaster-recovery(BC/DR) planning--until the introduction of tape pipelining.

Tape pipelining virtually eliminates the impact of latency on the sustainable throughput of tape backup over distance. With tape pipelining, the remote tape backup system appears local to the server and is able to sustain high performance over thousands of miles. This greatly enhanced performance over distance allows considerable flexibility to companies considering tape as part of their BC/DR strategy. Tape pipelining can enable a single central site to act as a backup facility for many geographically dispersed data centers. It can also enable backup from a central site to a remote site.

Companies can backup stored tape data across low-cost and readily available IP connections, using both newer native Fibre Channel tape drives as well as older SCSI drives. In addition, open systems servers such as UNIX and NT can be incorporated into remote tape strategies.

Many companies are using IP networks for remote disk mirroring which is typically synchronous and therefore uses substantial network bandwidth during the day. Since tape backup is typically asynchronous and often done during nights and weekends, tape pipelining means that businesses can now use their existing IP networks for both disk mirroring and remote tape backup solutions.

The result is improved data protection and recoverability in the event of a disaster and a significant reduction in the total cost of ownership of a remote BC/DR solution.

With tape pipelining, a single location can act as the backup for numerous remote servers and centrally manage the backup/restore operation. Tape pipelining over IP allows this to be done over plentiful yet inexpensive IP connections.

Selecting BC/DR Solutions

In planning BC/DR systems, businesses can select from a wide range of data-protection solutions (see Figure 1). The choices they make will be determined by the importance of the information they need to protect (i.e. its time sensitivity and business, value) and the amount of money they are willing to spend to protect it. These factors help determine the appropriate recovery lime objective (RTO).

Industry research shows that a typical mid-sized to large enterprise might have 150 mission- and time-critical applications that require continuous availability; 300 or so applications that require data restoration within four to 24 hours; and 100 or so non-critical applications that can wait as long as several weeks for data restoration.

Levels of information protection: If a business cannot continue profitable operations without its data, it usually implements a remote hot site that completely duplicates critical data and applications and makes them continuously available. The cost of such a solution is justified by the cost to the business of even a few minutes of down time. At the other end of the spectrum is business data that may not need to be restored for days, weeks or even longer after a data center disruption. In fact, some business data requires no restoration at all but simply needs to be protected for legal and archival purposes.

The most cost-effective enterprise-wide BC/DR plan will probably involve a mixture of disk and tape technologies, from the most expensive for data that needs to be continuously available to less expensive technologies for data of less time sensitivity and importance. Figure 2 shows what such a plan might involve.

New factors and increased focus in BC/DR planning: BC/DR strategies are affected by the increasing amount of business data being pushed into the mission-critical category, requiring continuous availability. Businesses are also aware that the dangers to their mission-critical data are greater than ever before. Not so long ago businesses had to plan for human error, technical glitches and, of course, natural disasters. Post 9/11, businesses must also plan for terrorist attacks that, as a November 2002 report from Illuminata put it, could "take out more than just a couple of floors of one data center."

Indeed, government agencies and private disaster planners are urging businesses to establish BC/DR systems that separate primary and backup sites by at least 200 miles, instead of the typical 60 miles. The SEC and Federal Reserve are calling for businesses that play a significant role in financial markets to establish fully redundant backup sites at least 300 miles offsite and be able to recover critical activities on the same business day of a disaster.

Tape Backup Market

The evolution toward networked storage, the heightened awareness of business-continuity and disaster-recovery planning and the need to reduce costs and improve efficiencies throughout IT, is stimulating the market for cost-effective tape and enterprise backup solutions.

IDC (September 2001) sees the tape market moving from solutions targeting the high end of the data storage market to lower-cost solutions that add more capacity than standalone tape drives. Through 2003/2004, as data volumes increase 200%-300% per year, Meta Group (October 2001) projected that backup/restore windows will be reduced using intelligent storage, highly parallel tape solutions, and FC-based SANs.

Gartner Group stated unequivocally (March 2001) that disk-based replication will not replace tape backup. Tape's low cost--it is about 10 to 60 times less expensive than disk storage--makes it a compelling economic argument for its inclusion in any high-availability storage infrastructure.

Tape Backup and Its Historical Limitations

Tape backup plays a big role in cost-effective BC/DR systems since it is significantly less expensive than disk-based replication and is perfectly appropriate for protecting many types of business information. However, historically there have been inherent performance problems with tape backup over distance that limited its role in BC/DR planning.

The primary reason for this limited role is the sequential nature of tape I/O operations, which makes tape backup highly sensitive to the latency that accompanies distance. Unlike disk-based operations in which blocks can be written in parallel, tape I/O for a single block must complete before the next block can be written. Figure 3 illustrates the typical tape performance falloff with distance for several popular tape products. As the chart makes plain, remote tape backup systems simply don't have the throughput to move large amounts of data over distances of more than a few miles.

Tape Pipelining Removes Distance Limitations

With tape pipelining technology, tape backup can play a larger role in BC/DR plans, even at distances of thousands of miles or more.

What is tape pipelining? Tape pipelining is a data-streaming technique for remote tape applications. It maximizes the efficient flow of data by using the concepts of buffering and error recovery, and extending these capabilities over remote distances via emulation.

How does tape pipelining work? In tape pipelining, the server and tape controller can sit thousands of miles apart, but the two units believe they are sitting next to each other in the same data center. Both the server and the tape controller have an intermediate device such as a storage router nearby that emulates the distant server or tape controller.

The exchange of information between the server and tape controller is buffered, allowing write/confirm operations to take place in tandem using temporary memory. For error recovery purposes, operations (and associated data) in the queue are retained until a successful completion status has been received from the controller.

Tape Pipelining Over IP

Tape pipelining over IP allows businesses to perform remote tape backups during those periods when bandwidth is essentially free.

For example, using existing IP networks between geographically separated facilities, companies can support their daily production network and real-time synchronous data replication and disk mirroring applications by day, and perform scheduled, asynchronous tape backups during evenings and weekends, when excess bandwidth is available.

Studies show that up to 80% of the total costs for a remote replication solution over five years are network bandwidth costs. Re-using existing bandwidth rather than maintaining dedicated lines can dramatically lower overall cost of ownership.

Key benefits of tape Pipelining over IP:

Businesses can use existing storage resources and IP networks as part of their BC/DR solutions by having a central site act as the backup for a number of geographically dispersed facilities.

Businesses can streamline server backups into one remote tape backup system because open systems servers, such as UNIX and NT, can be incorporated into previously mainframe-only networks.

Businesses can re-use existing tape drives by incorporating both the newer native Fibre Channel tape drives as well as older SCSI drives into the remote tape backup system.

Businesses can reduce the number of people required by centrally managing tape backup/restore operations.

Businesses can use existing IP networks and bandwidth, particularly during evenings and weekends.

Remote Tape Backup--a Key Part of BC/DR Planning

With support for tape pipelining over IP networks, companies can perform remote backups using readily available, less expensive IP networks. It also allows highly distributed organizations to centrally manage the backup operation for its distributed facilities, improving data availability of its distributed, mission-critical data (see Figure 4).

For many companies, analysts recommend that BC/DR plans provide disk backup (data replication or data mirroring) to secondary backup centers hundreds of miles from primary sites for critical information that needs to be continuously available.

Analysts also advise that other types of information be available within hours of a manmade or natural disaster. The most cost effective way to meet both needs is to use the same network infrastructure and remote facilities for both synchronous disk backup (for business information essential to continuous operations) and for asynchronous tape backup (for less time-critical but still important data).

Tape pipelining makes it possible by eliminating the latency effects that traditionally have accompanied tape backup over distance. Devices that make tape pipelining possible use buffering, emulation and advanced compression techniques to make local and remote controllers and servers act as if they were right next to each other.

This not only makes it easier for more businesses to use disk and tape together to provide maximum coverage for information based on business requirements and budgetary restrictions, it also allows businesses to take better advantage of their IP bandwidth.

With tape pipelining, remote tape backup/restore comes of age. For businesses seeking to improve their data availability while at the same time controlling costs by using existing hardware, software and network bandwidth, remote tape backup over IP makes perfect sense.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

Patty Barkley is storage networking marketing manager at CNT (Minneapolis, Minn.)

WWW.cnt.com