Despite pressure from an increasing number of disk-based backup products, tape remains the cornerstone of nearly all organizations' backup plans.
By Heidi Biggar
When it comes to change, IT administrators are notoriously resistant.
Consider the adoption rate of Fibre Channel storage area networks (SANs): Depending on which survey you read, only 20% to 25% of end users currently have them installed, though the technology has been available for several years. Direct-attached storage (DAS) is still the dominant storage architecture. Why? Because end users know DAS, are comfortable with it, and have invested too much money in it to justify an architectural switch.
What about iSCSI and IP storage? Despite its potential advantages over Fibre Channel, adoption has been much slower than most analysts had initially anticipated. Will end users embrace iSCSI? Most analysts say they will primarily because iSCSI has the potential to lower the barriers of entry to networked storage for many small and mid-sized companies. But the reality is that companies will likely implement iSCSI selectively at first (e.g., to centralize backup in a campus environment). In fact, Fibre Channel and iSCSI will likely coexist for years.
A similar transition is beginning to take place in the backup arena. The truth is that tape is facing new challenges from disk (see the following article, "Understanding disk-to-disk backup," p. 26), and in some cases disk may be a better choice than tape for backup-and-restore operations.
However, the reality is that tape—not disk—is the cornerstone of nearly every organization's existing backup plan, and it is the medium for which all major backup applications were originally engineered.
Does that mean that end users won't adopt disk for backup? Probably not. But what it does suggest is that, like everything in the storage industry, adoption will take place gradually. Rather than ripping out and replacing existing tape-based backup architectures with disk-based platforms, end users will likely implement disk-based alternatives for select data at first.
"Most of the mainstream vendors are [initially] focusing on using disk for staging to tape," says Jacob Farmer, chief technology officer at Cambridge Computer, a consulting firm specializing in backup. He believes that the ingredients of a scalable secondary storage system should include a mix of both disk and tape.
Dianne McAdam, a senior analyst with the Data Mobility Group, sees something similar happening. "I don't see tape going away as a backup medium," she says. "I see backup evolving into a two-tier architecture—with disk-to-disk backup satisfying the initial backup [and quicker restore] requirement, and tape taking care of the more permanent retention of weekly, monthly, quarterly, yearly, etc., data."
The purpose of this article, however, is not to debate the tape-versus-disk issue (and the related snapshot vs. mirroring replication arguments), but to take a look at the current state of the tape market.
Like many segments of the storage industry, the tape drive industry posted negative results in 2002. According to a November 2002 report from International Data Corp., the worldwide tape drive market was forecast to decline in both unit shipments and revenues last year, falling 15% and 23%, respectively.
IDC attributes the decline to several factors, including the overall slowdown in IT spending; the trend toward server/storage consolidation, which dramatically impacted sales of direct-attached tape drives (down 44% in 2002); and the increasing deployment of SANs, which favor the use of fewer, higher-capacity tape drives or libraries in a shared environment.
In the tape-vs.-disk debate, the segment of the market to watch is the midrange. Drives in this space are most commonly integrated into autoloaders and multi-drive libraries and then used in networked storage environments.
The midrange tape technologies that are in the spotlight include the so-called "super drives"—specifically, LTO 1/2, Super DLT 220/320 and, soon, Super-AIT.
This month, Sony is expected to begin shipping volume units of a 500GB Super-AIT drive. The drive, which is based on a 5.25-inch, not 3.5-inch, form factor, is not backward-compatible with the existing AIT format. The drive, however, does leverage many AIT technologies.
Elsewhere, media and library vendors continue to sign on as licensees of the LTO-2 format, and Hewlett-Packard began shipping limited quantities of second-generation drives to its OEM partners. Quantum, meanwhile, is shipping volume quantities of its second-generation SuperDLT drive—the SDLT 320 (see "At a glance" box, p. 22).
In addition to greater capacity and throughput, these "super drives" offer a variety of features (e.g., lower bit-error rates, higher bit densities, native Fibre Channel interfaces, etc.) that are designed to make the drives more reliable and efficient, as well as easier to use, in networked storage environments.
This activity underscores the reality that tape vendors understand the tenuous nature of their backup businesses, and they're responding by introducing increasingly powerful tape drives and increasingly intelligent tape libraries.
To remain competitive against disk, the Information Storage Industry Consortium (INSIC) says that tape areal density must grow at a rate comparable to disk areal density, while remaining competitive on a cost-per-megabyte basis. According to INSIC, that means a single tape cartridge must be capable of storing 1TB of uncompressed capacity by 2006.
If you ask analysts and vendors about the specific opportunities for tape in the coming year, you'll undoubtedly hear a lot of talk about regulations, iSCSI SANs, archival, and disaster recovery.
A new Health Insurance Portability & Accountability Act (HIPAA) privacy rule and the Sarbanes-Oxley Act of 2002 may well open new doors for tape—and disk. Though different in nature and scope, both regulations will affect the way in which the healthcare industry and corporate America retains data.
Enforceable as of April 14, 2003, the new HIPAA regulation establishes new guidelines for the transmission and retention of all health data and medical records. Specifically, the rule requires all healthcare and related companies to keep records (e.g., policies, patient requests and complaints) for a minimum of six years. A second, but not yet approved, rule concerning the security of this data is expected to specify electronic record retention periods.
Although neither HIPAA regulation dictates the type of storage medium (e.g., disk, tape, or optical) that an organization must use or where the data must be stored (on-site, off-site, etc.), it is likely that end users will store much of this data on tape or other backup media.
Similarly, the Sarbanes-Oxley Act dictates how corporations must store financial data, as well as any data used to compute their financial status. And it requires corporations to store information in a format that cannot be changed.
This has interesting implications for the evolving backup space—in particular, for the midrange tape market and certain types of new disk-based backup systems.
"Normal disk-to-disk backup solutions do not have the ability to guarantee that the contents of the file remains unaltered, but EMC's Centera does," explains the Data Mobility Group's McAdam.
McAdam says that while some disk vendors expect to see all backups going to disk, she doesn't see that happening—not now, anyway. "They are failing to consider all the data [e.g., e-mail, payroll/financial/health-care records] that, by law, must be kept for years," she says. "These are perfect applications for high-capacity tape."
As for iSCSI/IP SANs and disaster recovery and archival, the implications are more direct. "The low cost of IP SANs makes LAN-free backup much more affordable and [hence] keeps tape backup in the game," explains Cambridge Computer's Farmer.
IP SANs will not only make LAN-free backup a reality for many more users, but it will allow users to make better use of their existing tape resources (to consolidate and share tape resources over a common IP backbone) and will enable users to back up remote sites quickly, easily, and inexpensively—comparatively speaking—from a central location.
In fact, according to John Hufferd, the author of the recently published book, iSCSI: The Universal Storage Connection (Addison-Wesley, 2003), "backup to tape will be one of iSCSI's 'killer apps.' "
Other killer apps for tape are archival and disaster recovery. End users in all market segments will continue to invest in various tape technologies for these applications, as well as pure backup, in the months ahead.
International Data Corp. expects the worldwide tape market to continue to decline over the period from 2001 to 2006. Unit shipments are forecast to decrease 2% compounded annually, to about 2.4 million units in 2006, while OEM market value is expected to decline 5.9%, to about $1.9 billion.
While IDC expects a steep drop-off in unit shipments and revenue in the entry-level market segment, it does expect to see significant growth in the midrange market. The low end of the midrange market (tape drives with 40GB to 99.9GB of capacity), for example, will see close to 27% CAGR in unit shipments and a 2.5% increase in revenue over the forecast period.
Higher-capacity drive segments (>100GB) are also expected to show strong growth through 2006, buoyed by second-generation SDLT, LTO, and S-AIT drives, according to IDC.
The future of the tape backup market depends largely on the ability of midrange vendors to keep churning out higher-capacity, higher-performing drives, and the library vendors' ability to integrate them into products that meet evolving SAN and end-user storage requirements and price points.
At A Glance...Next-generation
Manufacturers: Hewlett-Packard (Ultrium 460 Tape Drive), IBM, Seagate
Availability: Limited quantities available from HP.
Availability: Volume shipments expected this month.
Manufacturers: Quantum, Tandberg
Availability: Shipping in volume.
HP, Seagate breathe life into DDS
Last month, Hewlett-Packard and Seagate announced that they haven't given up on DDS (digital data storage) tape—not yet, anyway. The two companies announced plans to develop, manufacture, and market a fifth-generation DDS drive this year, thereby breathing new life into a tape technology long buried by many.
Sony, one of the three original DDS manufacturers, has no plans to manufacture a fifth-generation product.
Despite competition from products from Benchmark (now Quantum), Ecrix (now Exabyte), OnStream, and Tandberg, DDS still owns the market, according to analysts.
HP and Seagate say end users can expect a significant boost in native capacity and transfer rate, as well as some "unique" solutions. The drives will be backward read-write compatible with previous DDS generations. DDS-4 has a native 20GB capacity and a 3MBps native transfer rate.