The Virtues of Virtual Tape

Posted on December 01, 1997


The Virtues of Virtual Tape

Potential benefits of virtual tape systems include increased performance and a sharp reduction in the number of tape cartridges, drives, and libraries required.

Flavio Santoni

Sutmyn Storage Corp.

Tape storage systems have consistently been a prominent feature in data centers for good reasons. Properly used tape media is by far the most cost-effective way to store data--even when figuring the costs of associated resources. Tape in MVS environments is compatible, removable, and transportable, providing essential portability for migration, backup, archive, distribution, and disaster recovery.

But the typical data center today is hard-pressed to keep up with the explosion of storage requirements fueled by the ever-increasing demand for more information from such applications as on-line transaction processing and data warehousing. As a result, massive tape libraries are quickly consuming the raised floor space reclaimed by new, smaller-footprint systems and high-capacity DASDs. Data-center managers are keenly interested in gaining control of growing tape resource expenditures.

Despite recent advancements in system and DASD performance, tape has been somewhat overlooked. Drive performance has improved, robotics have become more efficient, and cartridge capacity has increased from 200MB to 30GB. Nevertheless, most of the significant technology advancements have been focused on increasing the efficiency of other, more expensive storage resources.

The full potential of tape technology is not being realized by most users for a number of reasons, e.g., poor cartridge- capacity usage, insufficient use of tape system bandwidth, and mount delays.

While the cost of storing data on tape is declining and capacity is increasing, tape is not a cost-effective or efficient technology unless its capacity is more fully realized. Today, only 10% of available mainframe tape storage capacity is typically used, and cartridge capacity is growing faster than the amount of data actually stored in the cartridges. In fact, approximately 63% of a data center`s cartridges with single data sets contain less than 200MB of data, with an average of 50MB each (see Figure 1). Improved media density only serves to further increase costs and wasted space. Additionally, small files account for over 30% of the daily mount activity, which necessitates expensive robotics.

From the media cost standpoint, you can conceivably store data on 3490E cartridges at a cost of $0.002 per MB. However, if your cartridge only stores an average of 50MB of data and the cartridge costs five dollars, you`re actually paying $0.10 per MB. For data-center managers, that means expenditures for media are 50 times higher than they would be if the media were used efficiently. The total cost of ownership is higher still when you consider library (up to $50 per slot just for the robotics) and floor-space ($300 per square foot) costs--not to mention the expense of other related resources, e.g., drives, controllers, and annual maintenance. By using 80% to 90% of 3490E cartridge capacity, overall costs will drop dramatically.

Similarly, most applications use only a fraction of the performance potential of tape drives. Although tape-drive performance has improved dramatically, over 60% of tape applications transfer data at a mere 0.5MBps (see Figure 2), which is just a fraction of the drive`s performance capabilities (6MBps to 18MBps with 2:1 compression). Tape applications require more tape drives, which means that the only way to run more applications in parallel is to buy more drives, further wasting bandwidth and money.

Other tape problems include mount delays and bottlenecks. Tape-drive mounts are mechanical, and physical mounting takes time. Even the fastest robotics mechanisms and transports still take 25 to 40 seconds to mount and ready tapes. In a production environment, these delays are detrimental to the batch-processing window. Batch processing needs to be completed within a specific time frame so that inquiry applications can be brought on-line as quickly as possible.

All of these drawbacks constitute inefficiency, coupled with ever-increasing requirements to reduce processing time, they create enormous time and cost problems.

The Solution: Virtual Tape

Storage experts are aware of the causes of these inefficiencies and how the natural bottleneck of human- or robot-mounted tape makes them worse. This understanding has led engineers to create software and hardware solutions that bring disk-like access times and performance characteristics to tape processing, while significantly improving media usage. As a bonus, the same solutions that increase storage capacity usage and shorten retrieval time can also save CPU cycles by taking intelligence outboard to provide faster throughput.

Virtual tape is regarded by many analysts as a revolutionary storage solution that may change the tape storage market in the same way that the "virtualization" of operating system memory and RAID has affected information processing. Virtual tape systems transparently provide instant access to data sets during active periods and to low-cost removable storage for archiving during inactive periods.

For example, some virtual tape systems can create an image of 32 drives and 100,000 volumes in less than 10 square feet, providing up to 100GB of effective tape cache. They can reduce current cartridge requirements by more than 35%, maximize tape bandwidth usage, reduce the number of required tape drives, and provide the benefits of faster and higher-capacity technologies without expensive and time-consuming conversion efforts.

How Virtual Tape Works

A virtual tape system emulates a group of standard tape drives. Based on specific criteria, data sets are directed by software to the virtual tape system. Initially, the small and medium data sets that cause cost and efficiency problems by creating significant tape mount activity and by using unnecessarily large amounts of robotics or manpower resources are directed to the virtual tape system. These data sets usually remain active for only a short period of time after creation. During this period, when multiple recalls may occur, the data sets are kept in tape cache--RAID disks--to optimize mount performance. Access to these data sets occurs at disk-like speed throughout the period.

By using virtual tape technology, thousands of mounts per hour can be performed, each in under a second. This is a quantum leap over current robotics systems, which typically perform a single mount-to-ready in about 30 seconds and are only capable of a few hundred mounts per hour.

After the active period, the likelihood that the data sets will be used again is dramatically reduced. Therefore, the data sets are removed from the virtual tape cache and efficiently stacked on various media types, using existing equipment to maximize media usage. The appropriate catalog information is updated accordingly, and cartridge capacity is maximized. Studies show virtual tape will enable users to use some 35% fewer tape cartridges (see Figure 3).

Vendors have different approaches to implementing virtual tape. Some solutions require proprietary libraries, while others optimize existing libraries or standalone drives.

What Users and Analysts Say

"Virtual tape is the advance in tape management that we`ve been looking for," says Ken Harper, business unit leader at Acxiom CDC, a data processing outsourcing firm in Chicago. "Our processing of market data located on multiple tapes in multiple regions needs really quick turnaround, and virtual tape has given us the potential for great response time in tape batch processing. For example, we ran a test and experienced approximately 400 mounts per hour on traditional tapes. With virtual tape, we achieved 5,888 mounts per hour. The fundamental considerations that initially led us to virtual tape were environmental costs and space constraints," explains Harper.

Acxiom has 460 free-standing tape drives and had effectively run out of room. "Virtual tape offered us a cost- effective alternative to installing more tape storage. Virtual tape is a win for everybody," says Dan Kaberon, manager of computer resource management for Hewitt Associates, a provider of outsourcing solutions for benefit plan administration, in Lincolnshire, IL.

"We use it to route allocations of small tapes, and we get better performance and less media consumption," explains Kaberon. "There are real cost issues related to users who are pathologically inefficient in their use of tape. With virtual tape we can fix this without making JCL changes or any changes at all for the users."

But that`s not the only advantage of virtual tape at Hewitt. "There is also a big secondary gain," says Kaberon. "A reduction in tape usage means a big reduction in reliance on robotics, which makes our current robotics configurations more effective. With fewer things running on them, the robotics perform better, which improves our overall system`s effectiveness."

A virtual tape system can be installed without changing the JCL or creating an ACS routine, and it can be managed using available library management software. The software runs on MVS and interfaces to most popular tape management systems, including Computer Associates` TMS and TLMS software and IBM`s RMM, and typically, virtual tape systems can be installed in one day without disruption.

Virtual tape systems are currently compatible with System/390; next year, they will work with open systems. Some are compatible with any tape or automated tape library system installed. This obviates the need for new tape drives and cartridge media technology, thus protecting users` investments in existing tape/robotic hardware assets.

John Young, vice president at The Clipper Group consulting firm, in Wellesley, MA, says, "Virtual tape systems will have a positive effect on the cost of System/390 computing, comparable to the evolution of CMOS processor technology and RAID DASD architectures. Enterprises using virtual tape technology will substantially lower their investments in tapes, tape drives, and libraries; they will also reduce the associated environmental costs and the personnel costs required to manage the new environment."

"By mid-1998," he continues, "virtual tape technology is likely to embrace open-systems storage. Having this technology evolve from S/390 assures open-systems users of iron-clad support and assures the enterprise of further functional value for open systems and additional cost reduction."

"Virtual tape enhances existing media and makes it more effective," says Carl Greiner, vice president and director at the META Group consulting firm in Stamford, CT. "It currently has the greatest applicability in mainframe systems, but we`ll see virtual tape implemented across the board for desktops and small servers as well," he predicts. "This is the first iteration of a new technology, but it shows great promise and it`s giving new life to tape."

Paul Wolfstaetter, vice president at the Gartner Group consulting firm, in Stamford, CT, says that "with the cornucopia of operational hassles endured for decades by tape users, it`s little wonder tape technology has been the medium that users love to hate. However, with the availability of virtual tape subsystems, tape is on the brink of a renaissance. No longer will the historic paradigm of one tape cartridge equating to one tape volume, or a tape mount equating to a tape transport, be a truism. Of even greater significance, requirements for tape library space will be slashed by more than 10-fold."

So far, however, no standards exist for virtual tape. "As virtual tape technology evolves and expands across platforms and offers a full spectrum of features and functions, the question of standards must be addressed," says Wolfstaetter. "It is imperative that users apply pressure to the virtual tape vendors and demand industry-wide standards."

Virtual tape delivers a quantum leap in price/performance levels and functionality, while addressing the inefficiencies that plague today`s tape processing. Virtual tape systems translate into dramatic benefits that will enable business opportunities for enterprise computing while allowing data centers to preserve existing investments in tape drives, media, and libraries. The benefits include improved cartridge usage, improved performance, and reduced costs. In addition, virtual tape solutions promise enhancements that address tape`s traditional processing drawbacks and that provide users with a powerful option to store data on-line at faster speeds and lower costs.

Click here to enlarge image

Click here to enlarge image

Figure 1: Approximately 63% of data-center tape cartridges with single data sets contain less than 200MB of data.

Click here to enlarge image

Figure 2: About 64% of data-center tape applications transfer data at less than 0.5MBps.

Click here to enlarge image

Figure 3: Virtual tape systems significantly reduce the number of tape cartridges required. (Figures for 1997 are based on an actual site implementation; 1998 to 2001 numbers are based on modeling estimates.)

Click here to enlarge image

In Sutmyn`s implementation, Library Management Software (LMS) running on a mainframe automatically directs designated data sets to the Scimitar virtual tape system.

Flavio Santoni is executive vice president at Sutmyn Storage Corp. in Santa Clara, CA.

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