To decrease the heavy costs of "rip-and-replace" tape technology upgrades, try to preserve your existing investments.
By Jack Fegreus
Digital data storage (DDS)-better known as 4mm DAT-is the highest-volume tape technology of choice for the PC server market. The performance of DDS technology, however, has significantly lagged behind its high-end competitors. DDS-3, with a native transfer rate of 1MBps and an uncompressed cartridge capacity of 12GB, left a lot to be desired when compared to the performance of Sony's AIT, Exabyte's Mammoth, and Quantum's DLT.
The new version, DDS-4, boasts a native transfer rate of 3MBps and an uncompressed cartridge capacity of 20GB. These specifications place the drive far more in line with the storage configurations of servers now being sold to small- and medium-sized businesses. Assuming a nominal 2-to-1 data-compression ratio, DDS-4 drives and tapes provide the lowest-cost devices and media to back up 40GB of data.
Data Storage magazine (a sister publication of InfoStor) benchmarked an HP DAT40e external DDS-4 drive. (Sony also builds DDS-4 drives.) The HP DAT40e is an Ultra-Wide SCSI device that will work with SCSI-2 or SCSI-3 controllers in either single-ended or low-voltage differential (LVD) configurations. We tested the unit with a single-ended Adaptec 2940UW SCSI controller in a Dell OptiPlex Gxpro system running Windows 2000 RC3.
To test the HP DAT40e, we ran the Nova Technica Tape Benchmark v1.0. This benchmark allocates a large block of memory from which it streams data to the device. By streaming directly from memory, the benchmark eliminates bus bandwidth contention with other devices. The data can be streamed in block sizes of 2n KB, where "n" ranges from 0 to 6. Since most backup packages now use 64KB transfers, we limited our tests to this size block.
The benchmark also generates two types of data stream: purely random data and data that falls into a preset frequency pattern. The purely random data is not compressible and therefore provides a baseline for native throughput. When run with compression turned on at the tape drive, an uncompressible random data stream also provides a worst-case scenario for the tape drive. The drive attempts to compress this data-which is typical of highly compressed JPEG, ZIP, and e-mail archives-and in so doing, throughput will degrade.
The patterned data stream was modeled using Exabyte Mammoth and Quantum DLT7000 tape drives. This data stream consistently produces a compression ratio on the order of 1.9 to 2.1 on earlier-generation tape devices. With this in mind, we used 3GB data sets in all of our tests in order to reach a repeatable steady state in terms of data compression and throughput.
In all of our tests, the 4mm DAT drive compared quite favorably to a high-end 8mm Sony AIT-1 tape drive. The Nova Technica Tape Benchmark pegged the base throughput of the HP DAT40e at 2.87MBps.
In contrast, the base throughput of the Sony S300C AIT-1 drive was clocked at 2.95MBps. Nonetheless, when we ran the compressible data stream, throughput on the HP DAT40e soared by a factor of 2.66 times to 7.62MBps. While we never got to this level of compression when running a set of actual disk backups, compression and throughput remained consistently higher on the HP DAT40e in every test.
Since the DDS-4 design uses the same DCLZ compression algorithm found in DDS-3 drives, the extraordinary performance exhibited on the highly-compressible patterned data stream was traced to improvements in the SCSI firmware and larger buffers in the compression engine chain. Using a firmware-based "intelligent disconnect," the DDS-4 drive optimizes the time to burst data at 40MBps into a mini-buffer that is used to hold the data before the compression engine.
An 8MB buffer then holds up to 20 compressed DDS "groups" of data. The compression chain is optimized so that should the drive be given an opportunity to access the bus again, the compression and group buffering will be held off so that available data can be loaded into the drive and processed. As a result of the tuned SCSI bandwidth utilization, the HP DAT40e can move data through the compression chain faster.
To get the threefold boost in performance over DDS-3, the rotation speed of the helical scan drum-which houses two read and two write heads located at 90° intervals-was tripled to 11,480 rpm. To provide better contact between the tape and the read/write heads on the drum at this higher rotation speed, grooves were machined on the drum to create a negative air-bearing surface. In addition, the preamplifier is mounted directly onto the rotating drum as close to the signal source as possible to maintain a good signal-to-noise ratio at higher transfer rates. Finally, drum assembly balancing was a top priority, as components on the edge of the printed circuit board have to cope with forces of around 2,200 gs.
To increase tape cartridge capacity, the width of DDS-4 tracks were narrowed to 6.1 mm from the 9.1 mm used by DDS-3. In addition, the DDS-4's tracks are written straighter and thus exhibit greater linearity. The length of the tape in a DDS-4 cartridge was increased to 150 meters from 125 meters, the thickness of the tape reduced, and the media formulated with smaller magnetic particles.
While narrower tracks are essential for increasing capacity, they also increase the likelihood of overlap between adjacent tracks, which produces a higher level of low-frequency noise that can impair the read channel's ability to accurately decode data from tape.
To help eliminate this problem, HP has introduced what they call Advanced Sequence Detection (ASD) for the first time on a DDS drive. This feature gives a tenfold improvement in error rate at the head/tape interface when reading data and further increases the drive's data integrity when restoring data.
To reliably recover data written to tape, DDS-3 technology uses an 8/10 encoding scheme: For every eight bits of user data sent to the tape drive, two extra bits are added so that 10 bits are actually recorded on the media. Over a long interval, the effect of positive and negative transitions produces a rolling sum dubbed the Digital Sum Variation (DSV), which enables the whole serial waveform of data to be analyzed; at any point on the waveform, the waveform conforms to one of seven different states.
As a recovery technique, DDS-3 drives use a Class 1 partial-response maximum-likelihood (PRML) recording technique in a Viterbi detector circuit. The maximum-likelihood detector uses previous bit patterns to determine what the current bit pattern should be.
To this scheme, HP has added a trellis decision-making network for part of the 8/10 encode/decode rules and implemented the trellis in the DDS-4 hardware. As a result, the maximum-likelihood trellis detector is able to determine the most likely received signal sequence-at a rate of 108 million symbols per second. In effect, this adds another level of error-correction code to DDS-4. It also helps to accurately determine what a bit pattern should look like when normally occurring higher levels of low-frequency noise take the encoding envelope out of its prescribed digital sum variation level and would typically cause a decoding error. In theory, ASD has the potential to increase the base-level error detection by a factor of 10.
The HP DDS-4 DAT drive proved to have a base performance on a par with an 8mm Sony AIT-1 drive. Nonetheless, the DDS-4 drive showed consistent superiority in its ability to compress data and effectively provide significantly higher throughput writing and reading data.
A number of technical advances target the rotating drum, which contains the read and write heads. Rotational speed has been tripled to 11,480 rpm. The preamplifier is mounted directly onto the drum to put it as close to the signal source as possible. Grooves are precisely cut in the drum to create an air-bearing surface to eliminate signal deterioration with the higher rotation speed. Finally, the drum is carefully balanced using weight pads to ensure rotational accuracy.