Tape virtualization has many faces

Posted on April 01, 2006


Understanding the different choices you have is key to determining which technology best fits your needs, and in what situations.

By David Hill

After receding for a few years, virtualization is back in vogue in the IT industry. All kinds of virtualization-server, disk, and tape-are becoming more popular. The concept sounds simple: Virtualization is the logical abstraction of physical resources and hides back-end complexity (e.g., it can allow you to add disk or tape resources without having to bring an application down).

Trying to understand when and where to use virtualization sometimes causes confusion, and that is true in the case of tape virtualization. The confusion starts because the word “tape” is used loosely to refer to the physical tape cartridge as a piece of media on which data is stored, the tape drive into which a piece of media is inserted, and/or a tape library, which contains tape drives, slots, and media.

The confusion continues because disk drives are often used to augment or complement tape virtualization. The result of this loose nomenclature is that it is not clear what is being virtualized (tape, drive, or library) and what is used to implement virtualization (tape or disk, or both).

Tape virtualization in a nutshell

To avoid this confusion, consider the choices available for tape virtualization:

  • Virtual tape-Disk is used as a cache to concatenate datasets in a manner that most efficiently uses the capacity of a tape cartridge.
  • Virtual tape library (VTL)-Disk is used to emulate a physical tape library.
  • Tape library virtualization-Tape drives and tape slots can be allocated dynamically rather than having fixed assignments.

Technically, writing to disk as a single virtual tape drive is a fourth alternative. That option may be appropriate for smaller configurations, but is not easily scalable to larger configurations.

Each of the tape virtualization choices solves a different problem (see table).

Click here to enlarge image

Virtual tape has been used primarily for writing mainframe datasets so that more of the capacity of tape cartridges is used. However, with the increase in size of open systems tapes, there may be a need to consider virtual tape technology for open systems as well.

The focus of a VTL is to improve the backup/restore process by using disk as a replacement for a tape library.

Tape library virtualization is still an emerging technology and refers to the implementation of virtualization software intelligence in the tape library itself to improve the sharing of both tape drives and slots.

IT administrators should first determine what problem they need to solve and then apply the appropriate tool. A closer look at each type of tape virtualization should help you in that decision-making process.

Virtual tape

Virtual tape refers to the virtualization of a piece of media rather than the virtualization of the tape drives in a tape library. Virtual tape has a longer history than VTLs and is commonplace in mainframe environments. On mainframes, the process of writing datasets to tape often left the tapes with a lot of empty space. With virtual tape, multiple datasets are concatenated on disk and then written to tape. Open systems typically have not had the same issue with empty space, but some efficiencies can still be achieved with virtual tape, and the technology is now available for open systems as well as mainframes.

The key advantages of virtual tape are improved asset utilization and ease of management. Virtual tape also achieves indirect benefits for data protection by minimizing the number of tapes that have to be restored, which leads to fewer chances for restoration problems.

Virtual tape libraries

Standard backup/restore software packages can target disk as well as tape. However, simply re-targeting standard backup/restore software from disk to tape requires that each backup job be manually retargeted to disk. That is not true of a VTL.

If the number of backup jobs that have to be changed is manageable, straight disk-based backup may be a feasible alternative. A second concern is that there might be, say, a 2TB file system limitation, which would apply to straight disk-based backup but not to a VTL. Thus, two potential benefits of VTLs in some situations are easy integration (saving administrator time and effort) and scalability.

A more-complex backup environment and/or high-capacity backup requirements (say, 6TB or greater, as a rough measure) would tend to favor a VTL. Otherwise, straight disk-based backup might be a reasonable choice, because it requires no additional software to maintain.

Tape library virtualization

Tape libraries tend to suffer from the same “sharing” problems that plague SANs with block-based storage virtualization. A SAN’s drives are partitioned so that each server can see a configured set of logical units. The drives are not “shared” in the sense that multiple servers can use the same logical units.

The same is true of tape drives and slots in a library. The allocation of servers to tape drives/slots tends to be fixed, with rigid partitioning. Administrators would like to create logical tape libraries that are larger than the number of physical libraries/drives that actually exist.

Dynamic partitioning enables storage slot pooling and flexible drive assignments. That, in turn, enables re-configuration of tape drives and storage slots as needed without having to take the library offline.

Obviously, only one application can use a tape drive at a particular time. However, use of the tape drive/slot can be “time-sliced,” with each server/application using the tape when another does not need it, allowing faster backup and restore via “parallelization” across tape drives. Sharing by fixed partition, in contrast, puts an added burden on the administrator to “guess” a priori the most-efficient partitioning, leading to reconfiguration problems down the road.

Putting storage virtualization intelligence in the tape library itself is not yet commonplace among tape library vendors. However, such intelligence can lead to not only more-efficient use of the tape library resources, but may also alleviate the need for a VTL because it could lead to more timely completion of backup jobs.

Getting the most out of your tape assets is important. Tape virtualization is the secret sauce that helps you do that, but you have to choose the flavor or flavors that best meet your needs. Virtual tape helps you get the most out of your tape media. A VTL can reduce backup times and improve the reliability of data restoration. Tape library virtualization can help you get the most out of a tape library. The choice is yours.

David Hill is principal of the Mesabi Group (www.mesabigroup.com), a consulting firm that specializes in storage networking and management. A version of this article originally appeared in the Pund-IT newsletter (www.pund-it.com).

Comment and Contribute
(Maximum characters: 1200). You have
characters left.

InfoStor Article Categories:

SAN - Storage Area Network   Disk Arrays
NAS - Network Attached Storage   Storage Blogs
Storage Management   Archived Issues
Backup and Recovery   Data Storage Archives