Tape capacities continue to spiral upward to the 1TB mark as speeds increase, but there's more to consider than just size and speed.
Although disk is rapidly becoming the preferred initial target and source for backups and restores, tape is still a key part of the process, especially for offsite disaster recovery and data archives. In responding to tape's changing role, vendors are enhancing their products to allow users to access tape-based data more quickly and securely.
Faster speeds and higher capacities are only part of the tape story: mile posts, cartridge memory, WORM formats and radio frequency identification (RFID) chips are among the options users need to bear in mind when evaluating the latest tape formats to determine what benefits they'll actually realize. Users deciding whether to hold onto what they have, upgrade or change tape formats should consider these key questions:
- Will the tape's larger capacities or faster transfer rates actually be used?
- Does data need to be encrypted and shipped offsite, and how long will it be stored?
- How much time will it take to move backup data from disk to tape and recover from tape to disk?
- Does data need to be stored in an unalterable format?
- What advantages do RFID tags and memory chips provide?
Capacity vs. speed
Native capacity--which today ranges from LTO-3's 400GB to the DLTtape S4's 800GB--is the first thing to consider when looking at new tape formats. For users with a lot of data to migrate from disk to tape, the time it takes to move the data and the ability to store large sets of data on a single tape are likely important factors, so higher capacity tape formats may be a good option. Approximately 18 months ago, Quantum Corp. noticed a trend where users opted for higher tape capacities over higher performance. As a result, Quantum changed its DLT product roadmap to emphasize increased capacity over speed. "We made a big [DLT] capacity bet," says Mark O'Malley, Quantum's manager, strategic marketing, storage devices.
|The leading tape formats|
|Click here for a comprehensive list of the leading tape formats (PDF).|
Quantum's new DLTtape S4 tape cartridge scales to 800GB native and 1.6TB compressed. DLTtape S4 provides a 167% increase in native capacity over its previous generation Super DLTtape II, and offers 300GB (native) more capacity than its nearest competitors, IBM Corp.'s 3592 Gen 2, Sun Microsystems Inc.'s StorageTek T10000 and SAIT-1 tape formats. The tradeoff with DLTtape S4 is that its 60MB/sec native speed doesn't measure up to the higher speeds that some of its competitors offer. For example, the 3592 Gen 2 tape format supports 100MB/sec native; the T10000 is rated at 120MB/sec native; and LTO-3's current speed is 80MB/sec, with a roadmap that says it will hit 160MB/sec native by the end of 2006.
|Pick the right tape format|
While there's no magic formula for picking exactly the right tape format for your organization, growing capacities, higher throughputs and different media types make some tape formats better suited for your environment than others. Here are some factors to consider:
Tape format tradeoffs
The choice of higher capacity vs. higher speed depends on a number of factors. In shops where tape is the primary target for backups, the flow of data from the server to the tape drive during backup may fluctuate, which can cause a backhitch (a repositioning of the tape by the tape drive). Tape formats like AIT and SAIT are less susceptible to backhitches because they use drives that lay the data down on tape using rotary heads, which makes the tape move slower. If the backup data flow is interrupted, it takes less time to reposition the tape than with linear tape formats, such as LTO or 3592, which rapidly lay down data by streaming the tape media through the tape drive.
Some tape drives that support linear tape formats, like IBM's TS1120 and Sun's StorageTek T10000, are taking steps to minimize the occurrence of backhitches. These two drives have a speed-matching feature that automatically changes the speed of the tape by sensing the incoming speed of the data. The TS1120 includes six options--50MB/sec, 60MB/sec, 70MB/sec, 80MB/sec, 90MB/sec and 100MB/sec--while the T10000 provides two speeds, 50MB/sec and 120MB/sec.
In shops deploying a disk-to-disk-to-tape backup process, the time restrictions to move data to tape are generally less constraining than when backing up data to disk. For users with longer windows to move the data from disk to tape--and who are less concerned about fast restore and access times--larger, slower moving tape formats like DLTtape S4 or SAIT may be the best choice. For instances where it's necessary to quickly move and restore data to and from tape, formats with higher native throughputs, like LTO or the StorageTek T10000, should be considered. Users employing tools like hierarchical storage management and accessing data on tape, may prefer 3592 tape formats that include "mile posts," markers on the tape at four-meter intervals that help pinpoint where the data lies. If data encryption is required, higher capacity tapes may be preferable because encryption eliminates the option of compressing the data.
|What's a 'super tape'?|
The term "super tape" is often applied to a group of tape formats, including 3592, 9940, AIT/SAIT, DLT/SDLT, LTO and StorageTek T10000, which are approaching 1TB native capacities. The following minimum characteristics roughly define the super tape category for 2006.
Native maximum capacity of 400GB. With disk becoming the primary target for backups, enterprise users can now take advantage of the larger tape capacities to minimize the number of tapes they manage.
Native maximum transfer rate of 50MB/sec. Data transfers from disk to tape can occur at faster transfer rates without errors.
Media form factor of 8mm or a half-inch. Larger media form factors of 8mm or a half-inch support the additional tracks required for higher capacities and transfer rates.
WORM option. The option to purchase or convert tapes so they store data in an unalterable format should be viewed as a necessity for users with compliance requirements.
Tape's actual cost
The final reason to use--or not to use--a higher capacity tape is financial. DLTtape S4 delivers roughly double the native capacity of SAIT-1 and LTO-3 for approximately the same price. However, buying tape based strictly on the cost per gigabyte doesn't take into account other cost-related issues, such as the need to upgrade tape libraries. Managing data on larger capacity tapes may also be a concern. "Capacity is getting so large that users can't use all of the space on the tape," says Richard Leonarz, marketing manager for tape storage in Sony's Media & Applications Solutions Division. Each succeeding generation in the "super tape" category becomes quicker and holds more data in a Sisyphean effort to keep up with growing data stores (see "What's a 'super tape'?").
But jamming more bits onto a tape introduces new compatibility and usage problems. LTO-3 tape drives can read and write LTO-2 and LTO-3 tape cartridges, but read only LTO-1 tape cartridges. SAIT-1 and StorageTek T10000 tape drives can only read and write to their respective SAIT-1 and T10000 tape cartridges; they can't read or write from older AIT, 9840 and 9940 tape formats. And as tape capacities and data transfer rates increase, users need to ensure that the tape drives supporting these formats can adjust tape transfer rates to match the slower speeds and feeds their infrastructure will likely produce.
Though relatively small, the demand for WORM tape is growing. Suppliers of LTO and SAIT say their WORM tape sales are between 2% and 5%, while Sony reports that sales of its SAIT WORM tape accounted for 7% to 9% of its total tape sales in the last six months.
Tape vendors implement WORM technology in very different ways. DLTtape S4 and SDLT 600 tape formats don't require a special tape cartridge to store data in a WORM format, and users can enable the WORM feature on existing tapes. Quantum's DLTIce software option puts an unalterable electronic key on each tape to ensure that data on the tape can't be altered. Once the key is applied to the tape, data may be written or appended to the tape, but not deleted or overwritten. The main downside to this approach is the lack of a WORM option for older DLT formats like DLT VS160 or SDLT 320. It may also be more difficult to keep track of how many tapes are reserved for WORM in your environment because WORM data and normal data are both stored on the same type of tape.
Other tape formats require special WORM-enabled tape cartridges. Users of Sun's 9840, 9940 and StorageTek T10000 tape formats need to purchase tapes with its StorageTek VolSafe option; these tapes may be used in any of Sun's tape drives, assuming the drives have the proper microcode to use the VolSafe option. LTO tape users may face a double whammy as only LTO-3 tapes support WORM; in addition to needing special WORM tapes, users may also need to upgrade to LTO-3 tape drives.
MIC and RFID
Vendors are starting to use memory in cartridge (MIC) chips and RFID tags more frequently, but these technologies come with their own issues. Almost all super tape cartridges contain MIC chips, except the DLT/SDLT families. While MIC chips are used primarily by tape libraries to track the location and placement of tape cartridges, in the future MIC chips will be used by applications for data classification, data searching and retrieval. However, the use of MIC chips today is limited by the lack of standards that define what data they contain and how that data should be stored in the cartridge.
Another concern is possible damage to the MIC. Although the possibility of damage is low, MICs can potentially store information such as tape indexes, WORM data, user data-access permissions, barcodes and media partitions. If the chip fails or is accidentally destroyed, the loss of data about the contents of the tape cartridge could be very troublesome.
RFID tagging is another technology whose promise is still unfulfilled; only the StorageTek T10000 includes this technology in its cartridges. Widespread adoption of RFID tags attached to tape cartridges may be driven by tape-vaulting companies such as Iron Mountain Inc., which are beginning to use RFID to track the shipment of tapes to and from user sites.
Tape remains a viable medium for long-term data storage and archiving. Yet as tape capacities and data transfer rates grow, and interoperability with previous generations of tape isn't always a given, some users may find that the latest generation of faster and fatter tapes is overkill for their environments. Before jumping to the next generation, carefully consider whether you need MICs, WORM and the additional capacity. For some applications and environments the answer may be "Not now."