A Tale of Two Tapes: The Many Challenges of DLT
With Ultrium products now appearing in the Linear Tape Open market, it is time to take a closer look at the technological advantages inherent in this format.
In November 1997, when Hewlett-Packard, IBM and Seagate Technology issued a joint press release that they were pursuing a new tape technology, the industry trade press gave the announcement coverage best described as minimalist. For several years, the "king of the hill" in the middle tier of the tape storage subsystems market – the only market segment showing significant signs of growth in the 1997-2000 timeframe – was Quantum Corporation's Digital Linear Tape (DLT). No one was having much success in displacing the leader.
Even when the Linear Tape Open (LTO) initiative, as the multi-vendor development effort was called, produced standards for two new tape products in April 1999, few in the media were quick to jump on the bandwagon. Against a backdrop of multiple announcements by established vendors, including Quantum, Exabyte and Sony, of improvements to existing tape products such as DLT Tape, Mammoth, and AIT, the announcement of not one, but two, new LTO tape standards, left many wondering why we needed any new tape technologies at all.
The initial value proposition of the new tape products was straightforward enough: With LTO products, customers could obtain a competitive tape technology from multiple sources. This contrasted with existing tape technologies, which were available only from a single vendor. LTO advocates argued that this situation exposed customers to potential disruptions in supply and left them at the mercy of single vendor pricing.
While a significant point, the press largely ignored the technical characteristics of Ultrium and Accelis, the two LTO formats. For the most part, the Ultrium product was described simply as "form factor compatible" to DLT, while Accelis was cursorily summarized as "a product derivative of IBM’s Magstar MP."
To Nick Harper, Vice President of Cybernetics, the appeal of Ultrium and other tape technologies derives from disenchantment with DLT. He says that a mixture of perceived and actual "ongoing operational problems" with DLT have already cost him lucrative accounts.
"Our first customer was lost because of perceived problems with the DLT solution we provided," Harper claims. "DLT tapes eat their leaders, especially if the cartridge is dropped and the clutch mechanism breaks." Whether or not tape reliability was actually an issue with DLT, says Harper, other problems reported by DLT users sent him looking for alternatives. "We have had numerous reports of DLT getting confused and losing directory information during use. When you do a skip search, the drive simply disappears."
Another often cited problem, according to the vendor, is heat generation, "DLT 7000 drives required considerably more ventilation in [computer and array cabinets] than other drives. 8000-series drives run even hotter and additional fans are needed to dissipate heat."
Another major drawback of DLT Tape, according to Harper, is the fact that media can be obtained only from a single source. "We see that as more important than a single source for drives. Most tape drives today are single source, except perhaps Travan."
Despite the numerous criticisms of DLT, and the fact that Harper does not expect reliability to improve based on Quantum's roadmap for the technology, Cybernetics still supplies DLT solutions when requested by consumers. "Selling DLT is the beginning of a masochistic relationship. We discourage customers from buying it, but if they want it, we will give it to them and support it. The consensus in the video world is that DLT [is not the best solution], but in the IT world, a lot of people have bought into it and continue to buy DLT drives and tape."
Ultrium Driven by Consumers
While it would be easy to cast Ultrium simply as a DLT competitor, it is more appropriately described as an entirely new approach to tape that integrates the strength of other tape products. In fact, few Ultrium vendors anticipate a mass migration away from DLT Tape to Ultrium in established DLT shops. Rather, they expect a long-term win for Ultrium as the limits to DLT and its successor SuperDLT are realized.
Ultrium developers performed considerable research in their efforts to prepare a standard for the tape. Interviews were conducted with users of a variety of tape products to discern what they liked and didn't like about existing products. From this process, several consistent themes began to emerge that supported the decision to move forward with product development.
One issue frequently expressed by consumers was the desire for an "open standard-based" product. This is an umbrella term used by Ultrium manufacturers to express several consumer issues.
In addition, many consumers verbalized concern about the failure of tape devices to keep pace with hard disk capacities, which were doubling every 18 months. DLT and Mammoth tape makers, for example, took six years to double their capacity from 20GB to 40GB.
Users were citing the limitations imposed by "Internet time" on the "window of opportunity" within processing schedules for performing backups, customers were concerned that the data transfer rates available from current generation tape devices would not enable backups to occur in the time permitted.
And finally, the metric that mattered most was restoral integrity. Consumers were deeply concerned that backups made to tape might not be recoverable when needed. Numerous highly-publicized reports underscored this point.
With these issues in mind, LTO developers set out to work on the Ultrium standard, envisioning an open format product developed by competing manufacturers (as opposed to manufacturing licensees) that would keep pace with capacity, transfer rate and reliability requirements expressed by prospective customers themselves. While the fact that the Ultrium specification described a drive form-factor compatible with DLT Tape suggests that vendors are seeking to displace DLT, vendors continue to claim that this decision is less about displacing DLT than preserving customer investments in tape library and autoloader technology. Ultrium is designed to support multi-mode library configurations that will emerge, vendors say, as DLT and other products "run out of runway" and newer Ultrium drives need to be substituted.
Ultrium Engineering Notes
Ultrium is not just another tape format, according to its designers. The specification combines the best-of-breed capabilities of existing tape formats with a set of clear-cut engineering decisions that provide not only a robust first generation product, but a reasonably clear roadmap for future generations.
Engineers have made several basic assumptions about tape technology that bear examination:
• Tape must double capacity and transfer rate every 24 months.
• Ultrium’s product roadmap reflects this position.
• A transfer rate of 10 to 20 MB/s is the minimum requirement to support current tape backup/restore operations.
• Access times must be less than one minute for a 100GB tape.
• High integrity for restoral must be guaranteed through sophisticated error checking and on-board historical data.
Looking at the HP offering, one can see how these criteria have guided the design of the first generation Ultrium products, which are now coming to market. The HP Ultrium offering features the critical components designed to meet the capacity and performance requirements sought by LTO designers, including eight magnetoresistive (MR) heads, coupled with eight digital peak detection read channels and a continuous timing-based servo technology. The selection of the older peak detection read channel technology may fly in the face of the current industry trend toward the use of Partial Response Maximum Likelihood (PRML) read channel technology with MR heads to achieve greater areal density (the number of bits of data that can be stored per square inch of recording media). Ultrium designers argue that the tried-and-true peak detection read channel is all that is required to meet current capacity and performance requirements for Ultrium and add that PRML will be implemented in the second generation product.
Of particular significance in this design is the 8-channel head, which writes data and two levels of Reed-Solomon error checking information across all 8 channels in parallel during write operations. HP has demonstrated that this capability delivers a significant integrity guarantee. Even if an entire track is damaged, it is still possible to recover all of the data from the tape.
Sub-minute access to data stored anywhere on the 100GB tape is provided through a combination of Ultrium’s continuous, timing-based servo and a cartridge memory. Cartridge memory, a repository for end-of-file marker and file directory information, has been a part of AIT tape from Sony Corporation for several years and is viewed as a significant value of that format. Ultrium takes the Memory-in-Cartridge (MIC) chip concept a step further, leveraging Phillips smart card technology to store not only recorded file layout information, but also other information that will be of enormous use in ensuring tape integrity. The cartridge memory provided on the HP Ultrium product stores data on – the number of times the tape cartridge has been mounted – facilitating the management and grooming of tapes based on service life; error rates associated with read/write operations – another indicator of tape life; and "generation of format" – information that identifies how many times the tape has been formatted and the date of the most recent format – again, useful in estimating tape life.
Ultrium’s cartridge memory also facilitates new technologies for tape management and library-based deployment. Memory can be scanned without inserting the tape into a drive and reading it. HP will soon be offering hand-held scanners and integrating scanning technology into automated tape libraries to facilitate high-speed selection and mounting of individual tape cartridges.
Longer Runway For Tape
Ultrium’s product roadmap envisions the expansion of the technology to per tape capacities of nearly 0.8 Terabytes within a decade. HP’s product rollout will provide a glimpse of things to come. The company plans to support both SCSI and Fibre Channel interconnects for its Ultrium drive-based products and refers to them as Storage Area Network (SAN) ready. A network attached storage (NAS) device integrating Ultrium tape technology is also in the works.
However, beyond the traditional roles for tape in backup/restore and archive, the performance characteristics evidenced by Ultrium lend further credence to vendor claims that tape may ultimately find new roles in non-traditional settings.
VTS was originally positioned as a vehicle for filling poorly used tape to capacity in order to reduce tape costs. The idea was to create a cache of disk drives to hold temporarily the data intended for writing to tape. Once a sufficient volume of data was cached to fill a tape cartridge completely, the cached data would be written to tape. The idea played well as a cost-savings measure in environments that were tape intensive.
Used in a web hosting environment, VTS takes on a different role. Large websites, those consisting of thousands of pages of HTML code, graphics and scripts, might be contained entirely on the tape components of the VTS. In operation, the most frequently accessed pages would be written to the disk cache, while less frequently accessed page elements would be obtained from tape when requested. Optimized for delivery at Internet speeds, such a strategy has potential. It would also reduce web hosting costs significantly.