Tape is still the best fit for applications that have lengthy backup windows, long-term data archiving needs or for organizations that require off-site storage on a medium that can be recovered using a variety of vendors' devices. For medium-sized and large enterprises, that usually means high-performance tape.
The key to taking advantage of improved tape technology is determining the appropriate media for specific applications. There are five major high-performance tape media on the market today: 3592, 9940, LTO, SAIT and SDLT. Selecting the right one begins with understanding the advantages of each. (See "LTO tape put to the test")
Mainframe environments will want to continue to use either 3592 or 9940 to satisfy their needs for batch processing, frequent file recalls and data managed under hierarchical storage management (HSM).
LTO, SAIT and SDLT formats will continue to lead in midrange shops. With a wide variety of cartridge manufacturers supporting these formats, tape prices remain competitive. Understanding each tape's features, unique technologies, application fits and ultimately their price-to-gigabyte ratio will help lead to the right tape media selection.
The performance kingpins
Five tape families meet the criteria for high-performance tape: IBM Corp.'s 3592, Storage Technology Corp.'s (StorageTek) 9940, LTO, SAIT and SDLT. The latest versions of these tapes offer 100+GB storage capacities, transfer rates exceeding 10MB/s, mean time between failures (MTBF) exceeding 400 days assuming continuous use and tape life spans exceeding 15 years (see "Tape specifics").
|The case for SAIT|
The AIT family of media is the only one of the five major tape media that puts data down on tape using a helical scan instead of linear placement. Helical scan places data on tape the same way a VCR stores data on a VHS tape: storing data across the entire width of the tape with a rotating head. The DLT and LTO linear process lays data down back and forth from one end of the tape to the other in multiple rows by streaming the tape through a stationary read/write head.
While the helical scan reduces backhitches, the case against AIT was that the spinning head knocked particles and debris off the tape, reducing its life span. To correct this problem, current AIT and SAIT cartridges ship with Advanced Metal Evaporated (AME) technology, which doesn't flake off as much as the older media did. Now AIT media has about the same longevity and reliability characteristics as linear tapes.
The other big change made with the release of SAIT was its shape. After AIT conquered its reliability problem, few users wanted to purchase new libraries to house AIT media. To address this, SAIT media has been retrofitted to the same external height, width and length characteristics as LTO tape cartridges, allowing them to fit into LTO slots in an LTO library. SAIT offers users immediate backup relief by eliminating or reducing backhitches and having the ability to use their existing LTO tape libraries without a significant capital investment.
The IBM 3592 and StorageTek 9940 are used predominantly in mainframe-centric shops. The IBM 3494 libraries, which host this tape type, are only managed under the mainframe operating system that manages data on tape just like it does on disk using HSM. The 3592 tape media is well-suited to the frequent start and stop nature of HSM environments, expediting small and selective file recalls as well as batch processing transactions predominantly found in the mainframe environment.
StorageTek's 9940 cartridge represents the other more mainframe-centric tape media solution. However, this media may also be considered for large Unix and Windows environments because StorageTek tape drives are supported by backup software that runs on these systems. The media offers benefits similar to 3592 media, but the 9940 offers a write once, read many (WORM) option not found on 3592 tape. StorageTek's latest 9940 VolSafe data cartridge provides non-erasable, non-rewritable WORM features for users who want to store data in a tamper-proof format. The knock on the 9940 as of late hasn't been the cartridge, but StorageTek's delayed mainframe FICON support for its tape libraries, resulting in some customers opting for IBM's 3494 tape libraries and the 3592 tape.
Of course, for a variety of reasons, LTO, SAIT and SDLT are the most commonly selected tape media families in Unix and Windows environments. First, all three are widely supported in many tape libraries from a number of vendors. ADIC, Exabyte, Hewlett-Packard (HP), IBM, Overland Storage, Plasmon, Qualstar, Quantum, Spectra Logic, StorageTek and others offer libraries that support these types of tape media. In the cases of ADIC, Quantum and StorageTek, they offer libraries that concurrently support media from the different families of products.
Unlike the 3592 and 9940 formats that are manufactured by a select number of vendors, LTO, SAIT and SDLT are manufactured by a variety of third-party sources. LTO best represents this ideal, with the standard cooperatively set by Certance, HP and IBM while the media itself is available from a variety of manufacturers. This standard enables IT shops to separate the buying decision for tape media and tape libraries and drives.
Every tape user asks: "How much will it hold and how fast will it go?" Yet paying more for faster, higher capacity cartridges may be pointless if users don't use the extra capacity during backups or there isn't enough time to store or recover the data. The more important question becomes, "How well can existing tape drives take advantage of these higher capacities?"
While faster tape speeds will correctly lead one to assume more data gets stored on tape more quickly, issues like backhitches--an interruption in the flow of data onto tape during a backup--have a big impact on the overall speed of the backup. However, tape drive vendors have taken steps to mask the backhitches by putting cache and more intelligence in the tape drive (see "Tape: alive and full of options," in the April 2003 issue of Storage).
This additional cache and intelligence in the tape drives becomes the key to maximizing both capacity and speed on newer tape cartridges. For instance, even though the latest 3592 tape offers up to 300GB of native capacity, it's only with advancements in the 3592 tape drive that the 3592 tape capacity and speed can be maximized. The 3592 tape drive enhances current backhitch countermeasures by incorporating a new technology called recursive accumulating backhitch flush into the tape drive. This technology allows the tape drive to respond to the random requests of backup software to verify that a tape write occurred without flushing its cache or necessitating a backhitch.
In addition to how well the tape and drive work together, drive prices also need to be factored into the overall tape backup solution. For example, expect to pay about $4,000 for an SDLT tape drive, $5,000 to $6,000 for an LTO tape drive and upwards of $30,000 for a 3592 tape drive. And then there's tape's so-called "infinite capacity." For years, mainframes have kept large amounts of older data accessible by managing it through their file system. But this isn't practical in most Unix and Windows environments. Their operating systems lack the sophistication to natively manage large amounts of tape.
Intelligent tape cartridges
Almost all storage devices are becoming more intelligent, and tape cartridges are no exception. The LTO and AIT tape families have for some time shipped with memory chips embedded in the cartridges with the most current tapes shipping with a radio frequency (RF) interface. LTO-Ultrium cartridges contain a 32KB RF-capable semiconductor while the AIT media families come with a similar RF-based 64KB memory chip. These cartridge-based chips store pertinent information about the tape, such as how many times the tape has been used, a log of the information stored on the cartridge and index information. This information becomes invaluable when moving tapes between different tape libraries or when used by different backup products.
RF chips also reduce wear and tear on both tape cartridges and tapes. Instead of physically handling the tape and inserting it into a reader, the RF signal transmits the information contained on the cartridge chip to a receiver either in the tape library or in any third-party device. This RF transmission is a cost-effective way to catalog and track tapes as they move around on site or off site.
However, the key for RF chip technology to succeed centers on the adoption of standards for all tape cartridges. For example, due to their different capacities, the LTO and AIT chips hold different amounts of data. They also store data on the tape in different formats (linear vs. helical) and use the chips differently to communicate to the tape drive how to retrieve data stored on the tapes. As a result, until RF-enabled chips become standardized across the most widely used tape formats, the ability to take advantage of this technology is limited. In addition, no third-party vendors provide readers for either LTO or SAIT RF technology and the DLT family of tapes has yet to incorporate a chip into their cartridges. In other words, buyer beware: This promising technology is something to watch at this stage, but not to be viewed as a product differentiator.
The embedded chip highlights current attempts by tape manufacturers to reduce the costs associated with managing tape. Accounting for where every tape is--whether it's in use, lost, damaged, purchased, expired, on-site, off-site or in the library--and what data is on it quickly eats up a lot of time and labor. It's a poorly kept secret that most companies never know where all of their tape cartridges are or what's on them.
Tape vendors are taking some steps to address this. The LTO and AIT RF-enabled chips and new management software, such as DLTSage for DLT media, should help alleviate some of this management pain and provide better reporting tools, but most companies are still struggling to achieve effective tape management. To do so, organizations must standardize on one family of tape and its most current release to take advantage of these tools. Even then, organizations must train personnel to develop proficiency on these tools. And data must be migrated from older tapes to these newer tape formats, a challenging and time-consuming task.
And of course, there's the tape-restore problem. Storage managers should carefully consider the applications they currently back up on tape, the time it takes to restore the data and the costs of exceeding the agreed-upon recovery time frames. While some administrators may still have the luxury of "best faith" recovery agreements (where their users accept whatever time it takes for them to recovery the data and get the application operational), they should no longer assume this will remain a given going forward.
There are just too many uncontrollable variables and hidden costs for storage managers to confidently state that they can recover data or control costs using only a tape backup solution. Until recently, there were few cost-effective alternatives for backup and recovery, but ATA and Fibre Channel (FC) modular arrays have forever changed the landscape. With the price points for ATA arrays affordable for almost any size organization and the uptime on FC modular arrays now acceptable by most data center standards, users should begin abandoning tape and look to ATA drives for production backup and recovery.
You should consider using tape as a long-term archiving solution. ATA disk fails to provide an off-site store of the data or offer the infinite storage capacity that tape does. Placing the initial backup on ATA disk creates a lengthy window to make a copy of the backup to tape and gives users the ability to take full advantage of current tape speeds and capacities.
A well-managed organization could eliminate many of its current backup and tape management headaches and dramatically cut its costs by properly balancing the deployment of ATA drives and tape technologies.