Quantum redefines DLT ‘value’

Posted on March 01, 2006

RssImageAltText

By giving DLT drives the software functionality of SDLT, Quantum puts the squeeze on DAT with the launch of the DLT-V4.

By Jack Fegreus

With the introduction of its fourth generation of DLT “value” tape drives, Quantum’s product road map has a simplified naming scheme. While the old SuperDLT (SDLT) moniker is gone, drive development will continue along two distinct lines: DLT-V, where “V” designates “value,” and DLT-S, where “S” stands for “super.” Add to that a number that represents the generation of the drive.

The new DLT-V4 is priced at less than $1,000 for an internal unit and has an optional Serial ATA (SATA) interface, which is rare among tape drives. This is very important for low-cost servers that target small to medium-sized business (SMB) applications, which often feature SATA as their default internal I/O architecture.

What’s more, the new drive can format standard DLTtape VS1 cartridges to store 160GB of uncompressed data, which translates into 320GB compressed at a nominal 2:1 ratio. This gives these cartridges the highest capacity and lowest cost-per-gigabyte in the SMB market.

Vis à vis DAT 72, it is difficult to devise a traditional TCO scenario for backup operations in which DLT-V4 technology is not superior. Traditional TCO analysis typically turns on two issues: whether backup operations are to be attended or unattended, and the number of cartridges required to complete the backup process. On both counts, DLT-V4 drives have an advantage over DAT 72.

First, DLT-V4 drives sport a near 4:1 advantage in backup throughput, on which labor costs for attended backup are directly dependent. Second, the DLT-V4 drives have a better than 4:1 edge in cartridge capacity, on which materials costs for tape cartridges and an autoloader are directly dependent.


While the native throughput rate of the DLT-V4 is 3x faster than the DAT 72, actual backup throughput was nearly 4x faster (left). Looking at this data normalized to the native throughput rate reveals that the DLT drive has a distinct edge in handling compression (rate).
Click here to enlarge image

At a site where a full backup involves 50GB or less of data, a DAT 72 drive will adequately fulfill the basic needs for backup. As the volume of data for a full backup begins to approach the 100GB level, however, a single DAT 72 cartridge can no longer store a complete backup and that will directly impact the TCO equation for an unattended-backup scenario. As the data volume continues to grow, the cost advantage of the DLT-V4 with its ability to put over 300GB of data on a single cartridge simply gets better.

There is, however, a new twist in backup that is missed by traditional TCO analysis. Legislative mandates are setting new rules for long-term data retention and are introducing constraints on the integrity of electronic records. As a result, a rising concern for backup accountability makes the most important component of the DLT-V4’s value proposition its support for all of the proactive management features of high-end SDLT drives. In particular, the DLT-V4 can lock a tape cartridge as a write-once, read-many (WORM) storage device via Quantum’s DLTIce software.

To examine the potential performance of the DLT-V4 in an SMB environment, openBench Labs conducted all tests on a typical SMB midrange server: an HP ML350 G3 running both Linux and Windows. For our tests, we used an external DLT-V4 with an Ultra160 SCSI interface. All tests were conducted using DLTtape VS1 cartridges. Since disk throughput is a significant factor in keeping today’s faster tape drives continuously streaming, we stored the openBench Labs data set for the backup tests on a SAN-based RAID-0 disk array connected over a 2Gbps Fibre Channel SAN. Using our oblFile benchmark, we pegged read throughput for our test set to be about 79MBps.

To ensure that disk throughput does not impinge upon the backup process, the source drive must be capable of sustaining a minimum I/O rate that is 3x to 4x greater than the native throughput rate of the tape drive.

To determine probable upper and lower bounds for throughput during backup-and-restore operations, we ran our oblTape benchmark. We streamed data to tape in very dense 256KB blocks. These data blocks were generated as either highly compressible (2:1) or non-compressible, and then intermixed in varying percentages.

The difference between the upper and lower bounds of throughput performance is typically about 3:1. The throughput range is greater than the 2:1 compression range because compression schemes add metadata about compressibility to the original data. When the original data cannot be compressed, the metadata is pure overhead. As a result, more bits are written to tape than are contained in the original data set. Those added bits reduce the perceived throughput rate-calculated using the number of bits read and not the number of bits written-below the native throughput specification for the drive.

That greater variation is also reflected in real-world backup performance. The throughput observed when writing data to tape is highly dependent on the characteristics of the data being sent to the drive and the ability of the drive’s electronics to handle fluctuations in data compressibility. During normal backup operations, differences in data compressibility from file to file make it more difficult to keep the drive’s buffer full, and that makes the drive prone to halting.

To test the capability of the drive’s electronics to keep it streaming under changing data input conditions, we generated heterogeneous streams of data that combined compressible and non-compressible data in varying proportions. Next, we verified the results of our synthetic benchmarks by running backup-and-restore tests with CA’s BrightStor ARCserve Backup for Linux v11.5. Using a 10GB data set, which is large enough to provide consistent results that were statistically valid, we ran a series of backup-and-restore operations. Our test data set contained a mix of mail folders, databases, documents, spreadsheets, slide shows, Web pages, and image files.

Using our oblTape benchmark, we determined effective native throughput by running the oblTape benchmark with hardware compression turned off. Results from this test pegged the effective native throughput for a Quantum DAT 72 drive at 3.3MBps and 9.9MBps for the DLT-V4. This performance-based stratification is even more apparent when the complete benchmark results are plotted over the full gamut of probable data-from completely non-compressible to highly compressible-along with the mean throughput for our 10GB backup-and-restore tests.

One of the more interesting aspects to surface within this data is a measurable difference in performance characteristics during backup-and-restore operations. When the backup results are overlaid on the benchmark curve, we find the DLT-V4 to be measurably more effective at compressing data in our backup test set. Insight into this difference in compression can be gained by normalizing the benchmark performance data to the effective native throughput rate determined for each drive.

From the normalized data, it is apparent that DLT-V4 provided better compression performance-either higher data compression or less metadata expansion-over 80% of our benchmark data spectrum. Most notably, the DAT 72 drive began expanding the total amount of data being written to tape once the amount of non-compressible data reached 60% of the test stream. In contrast, only when non-compressible data rose to 80% of the stream did the DTL-V4 drive expand the amount of data written to tape.


Monitoring data streamed from our SAN array during a backup with the DLT-V4 drive and DAT 72 drive shows that the array was able to push data at 6x the native rate of the DLT-V4 drive as compared to just 5x peak for the DAT 72.
Click here to enlarge image

The benchmarked compression efficiency of the DLT-V4 is very important when it comes to real-world backup performance. Highly compressed archives, such as Zip files, are not the only source of non-compressible data. Mail, database, and graphics applications frequently provide some level of compression within the application to help manage disk space. All of those applications are critical sources of business data targeted for backup.

That’s why our standard 10GB backup test set includes a mix of MS Outlook, MS Access, JPEG, and GIF data files. Running BrightStor ARCserve, we measured a higher average throughput performance boost using the DLT-V4 as compared to the DAT 72 drive. While the native throughput rate of the DLT-V4 drive is 3x that of the DAT 72 drive, our tests pegged average backup-and-restore performance to be 3.75x that of the DAT 72 drive

Surveys of SMB users have shown that the reliability of data-backup systems ranks second behind cost as a major concern. Reliability, however, is rapidly growing beyond readability and reproducibility to encompass the wider notion of media integrity. In a growing number of instances, restoring backup data is no longer sufficient: New legislative mandates on data integrity are making it necessary to demonstrate that the restored data is accurate.

The most draconian of these mandates affects IT at securities brokers and dealers. The Securities Exchange Commission (SEC) Rule 17a-4 requires that a secure copy of every transaction be kept available should the SEC audit the firm. In particular, records must be maintained on “non-alterable, non-erasable media.” This mandate dictates that firms retain all customer records, financial transactions, bank records, and buy/sell orders. This rule includes e-mail and instant messages.

Using DLT-V4 drives, any site can easily extend traditional backup operations to secure archival operations, while leveraging their investments in DLT infrastructure. Firmware in the new drive supports the use of DLTIce, Quantum’s WORM compliance component of the DLTSage tool suite with DLTtape VS1 media. This extends the potential use of the DLT-V4 for archival operations, which have fewer performance constraints. At $0.12 per gigabyte, DLTtape VS1 media provides a 25% cost advantage over Super DLTtape II and a 33% cost advantage over LTO3 WORM media.

Creating a WORM cartridge involves writing a unique electronic key that cannot be altered on each standard DLTtape cartridge that will be made WORM-compliant. This unique identifier creates a tamper-proof archive cartridge that meets stringent requirements for integrity protection while providing full accessibility for reliable duplication.

The state of the media when that key is implemented is of critical importance. Once turned into a WORM-compliant archive, the media can no longer be erased or reformatted. As a result, DLTIce functionality is best implemented within the tape formatting utilities of third-party backup packages. In this way, the backup package can format the media with its own proprietary scheme to handle the mix of data and metadata before it locks the cartridge with the WORM key.

The current version of Backup Exec, a copy of which ships with the drive, does not support DLTIce. That capability is planned for version 10.5, which will be released later this year. CA’s BrightStor ARCserve Backup already has the capability to support the DLT WORM scheme on SDLT 600. The current version of ARCserve, however, does not recognize the DLT-V4 as having WORM capabilities. We obtained a special patch for testing only. Like Backup Exec, this functionality should be available in a shipping version later this year.

In particular, CA integrates DLTIce support into ARCserve’s Quick Erase tape utility, which is used to recycle scratch tapes by destroying the label and directory while leaving the underlying data structure intact. As a result, whenever Quick Erase is used with a new tape to create a WORM cartridge, an administrator must first format the tape to create a DLTIce WORM cartridge that ARCserve can use in a backup operation.

Once a backup application has created a WORM cartridge, the DLTSage package is essential for checking on WORM compliance. The DLTIce module within DLTSage can be readily used to check the status of any cartridge and identify

  • Whether the DLTtape media is a DLTIce WORM cartridge;
  • The DLTIce signature (electronic key) placed on the cartridge; and
  • Whether anyone has attempted to tamper with the cartridge.

As the government and industry bodies continue to set increasingly stringent requirements for records retention, WORM support is becoming an expected tape drive feature.

With tape an accepted alternative to magneto-optical solutions, there is a growing consensus that WORM tape technology provides the benefits of low cost/gigabyte, scalability, and low capital investment costs through support of mainstream tape libraries, and long archive life-over 30 years.

Support for such WORM archiving and full preventive maintenance in a tape drive priced at just $999 for an internal unit and $1,099 for an external unit make the DLT-V4 tape drive unique for any market.

Jack Fegreus is technology director at Strategic Communications (www.stratcomm.com). He can be reached at jfegreus@stratcomm.com.


openBench Labs scenario

UNDER EXAMINATION

DLT-V4 tape drive

WHAT WE TESTED

    Quantum DLT-V4 tape drive
  • Ultra320 SCSI interface
  • Native (uncompressed) capacity: 160GB
  • Measured native throughput: 74MBps
  • DLTSage software
  • DLTIce WORM cartridge support
  • Drive and media statistics
  • Device diagnostic tests
  • Firmware update delivery

HOW WE TESTED

  • HP ProLiant ML350 G3 server
    • Two 2.2GHz Intel Xeon CPUs
    • 2GB ECC RAM
  • SuSE Linux 10.0
    • GNU Compiler Collection v4.0
  • CA BrightStor ARCserve
    • DLTIce support for SDLT 600

BENCHMARK

  • oblTape v2.0

KEY FINDINGS

  • Benchmark pegged the native throughput rate at 9.9MBps.
  • Uses current DLTtape VS-1 cartridges to store 160GB of uncompressed data.
  • Supports full DLTSage suite for preventive maintenance.
  • Includes DLTIce module for WORM compliance.


Comment and Contribute
(Maximum characters: 1200). You have
characters left.