How do you come to market so late—the clock is already counting down to the launch of LTO-3—and expect to carve out a strong market niche? The only way to do that is with a strong product edge. On that count, the Certance LTO-2 drive has an interesting technology edge that just might be strong enough. Nonetheless, an extra measure in marketing never hurts. Included with every Certance LTO-2 drive is a fully licensed copy of NetVault, and enterprise backup package from BakBone Software.

Historically, backup was one of those back-office IT operations that were regarded as little more than another item to check off at audit time. Given that low priority, it’s not hard to understand that, as the volume of stored data expanded and operations staff numbers were trimmed, IT managers thought what they were doing was a terribly creative solution at best and a low-risk gamble at worst to get around a backup window issue growing more and more intractable. The quick and dirty option at many sites: skip the verification pass.

On the plus side, not rereading the entire tape and checking its data against the disk truly cuts wall-clock time for a backup in half. That’s a significant productivity boost. What's more, if an individual backup tape does have a data corruption problem, it will probably be limited to a small region of the tape. In any event, the compelling rationale was that there should be a good incremental backup tape created within 24 hours of the missing data.

For today’s tape drives, there are three electronics modules where things can go wrong: a host DMA interface, a data compression engine, and a data formatter. An electronic hit—a static discharge for example—in one of these modules could corrupt the data before it is written to tape. Should such an event occur, the corrupted data would not be able to be read once it was written to tape.

The key to the SmartVerify scheme is the utilization of a Cyclic Redundancy Code (CRC) record, part of the LTO specification. The CRC algorithm treats that host-data block as a giant binary number, divides it by another fixed binary number, and makes the remainder the checksum. That checksum can then be used to ensure that the data has not been corrupted at any time in the future by simply repeating the calculation and comparing the remainder to the stored checksum.

In the case of Certance CL 400 drives, as soon as a block of data is transferred from the host bus adaptor to an LTO drive, a CRC calculation is performed on that data block. The resulting CRC checksum is then stored in a record descriptor, which is appended to the data block. In addition to the CRC checksum, that record descriptor includes the length of record and flag bits, which are used for data compression and partial record identification. The extended data block, which now includes the record descriptor, is placed in a ring buffer.

A ring buffer is used by all LTO drives to keep the drive streaming tape as optimally as possible. With today's faster tape transport speeds and greater bit densities, stopping and repositioning a tape represent an enormously expensive performance penalty. As a result, all tape manufacturers pay close attention to their electronics modules for maximizing data flow. One of the unique aspects of this module for LTO drives is the ability to sense an incoming stream of incompressible data and turn off the drive’s compression circuitry until it can again play an effective role. As a result, the lowest throughput level from an LTO drive will be equal to its native throughput.

The next module is the data compression engine. To protect data moving from the ring buffer to the compression engine, the CRC checksum for each data block is recomputed and compared to the value in the descriptor. If the new calculation does not match the stored checksum, an error is reported to the host.

Once inside the data compression engine, the CRC checksum is once again used to ensure the accuracy of the compression circuitry. The electronics for compression and decompression is independent, which means they can work in parallel. As a data block is being compressed, the previous data block is being decompressed. Once the data block is decompressed, the CDC checksum and length of the record are recalculated and matched against the values in the descriptor. If any problems are detected, compression is halted and an error is reported to the host.

In the final stage of processing before writing the data to tape, C1 and C2 Reed-Solomon Error Correction Codes (ECC) are generated and stored with the data in the same tape track. On all LTO drives, these codes are used to enable the drive’s read channel to employ a well-known digital signal processing technique: PRML-4 (Partial Response Maximal Likelihood).

PRML technology originated at NASA to read weak signals from satellites deep in space. In essence, the read channel compares the measured signal from the tape with a known waveform in order to correctly decode signals. In particular, the C1 ECC is used to recover from random errors caused by noise in the signal while the C2 ECC is used to recover from larger errors caused by physical defects in the tape, including the loss of an entire track as a result of a bad head.

Going one step further, Certance's SmartVerify electronics also utilizes both the C1 and C2 ECC to protect against data corruption during the write operation. As a result, the only way for data to be written to tape in a way that would prevent it from being correctly decoded when next read (i.e., during a restore) is for two separate modules to fail at precisely the same moment. Short of the whole drive dying, a scenario of two simultaneous failures is a virtual impossibility.

With all of that error-checking taking place, the question that immediately comes to mind is what this means for performance. To answer, openBench Labs examined throughput performance with homogeneous streams of compressible and incompressible data, as well as with heterogeneous streams of data in which we varied the percentage of compressible versus incompressible data.

Using homogeneous data provides a means to set upper and lower bounds for probable performance. On the other hand, heterogeneous data streams provide a way to test the capability of the drive’s electronics to sustain continuous data streaming.

As data becomes more variable in a backup scenario, a tape drive becomes prone toward halting, which necessitates repositioning the tape in order to resume writing data. For drives in the new super tape class, repositioning is especially costly in terms of the average throughput rate. As a result, performance measurements using data streams of varying compressibility are far more important for predicting real-world performance experiences.

Since our main goal was to examine the drive’s suitability for backup in an enterprise scenario, all tests were performed using benchmarks running in user mode in order to incur all of the operating system overhead that any enterprise application would incur. In addition, it was essential that our test hardware not present any throughput bottlenecks atypical of an enterprise-class data center.

Testing was done on an HP ML350 G3 server sporting dual Xeon CPUs, DDR memory, and PCI-X expansion slots. For our tests, the server ran SUSE LINUX version 9. We installed an Adaptec Ultra320 SCSI host bus adapter that was attached to 4 Maxtor Ultra320 Atlas drives in a RAID 0 stripe set for Disk I/O. For tape I/O we installed an Adaptec Ultra160 host bus adapter.

We used our oblTape benchmark with the CL 400 drive’s compression turned off to peg native throughput at 30 MB per second. We then used oblTape to generate a stream of synthetic data distributed in a pattern that was designed to produce a nominal 1.8-to-1 compression ratio on a DLT 7000 drive. This test produces an anticipated high-water mark for performance. Here we measured throughput at 56.2 MB per second.

For our worst-case scenario, we generated a stream of purely random (incompressible) data and left the drive’s compression on. This simulates the problem of backing up a highly compressed file: The drive can waste cycles attempting to compress incompressible data and write simultaneously. Thanks to the ability to turn data compression off when incompressible data is encountered, throughput during our worst-case scenario slipped to just 29.8 MB per second. In contrast, we pegged the native throughput of an HP Ultrium 460 drive at 29.5 MB per second and the upper and lower ends of the performance envelope at 53.6 and 29.2 MB per second. Interestingly, native throughput for the HP Ultrium 460 is rated at 30MB per second, while the Certance CL 400 is rated at 35MB per second.

By partitioning the data stream generated by the oblTape benchmark into 1,000 block segments and then interleaving these segments with compressible and incompressible blocks, we can stress the performance optimization scheme of any drive and measure its effects on overall throughput in a real-world situation in which heterogeneous data will have varying levels of compressibility. This is precisely the problem scenario that the LTO drive’s ring buffer is designed to solve. Using the raw performance measured over a range of compressibility ratios, probable backup performance can be projected given a reasonable estimate of the number and size of compressed data files within a given backup job.

These tests provided an important performance envelope for throughput. First and foremost, we could not measure anything close to a native transfer rate of 35MB per second, which would have put the Certance LTO-2 drive on a par with the Quantum SDLT 600. What we did measure was a solid 30MB per second which had a slight—1.7%—edge over the HP Ultrium 460. More interesting, the edge noticeably grew as our test data became more compressible. With less than 30% of the data being incompressible, the Certance edge was in the neighborhood of 5%-to-7%.

To validate our results, we installed NetVault 7.1, an enterprise backup package from BakBone Software. With the drive, Certance bundles a CD with NetVault 6.5.2 and includes a full license for one server with the Certance drive attached. In addition, the software  can be upgraded to release 7.1 on the BakBone Web site.

Once NetVault was installed, we performed a series of backup and restore operations using a 5-GB data set. Our test data contained a mix of files that included a large number of text data files, along with a mix of HTML and image files from a number of Web sites.

 NetVault was first introduced as a UNIX backup package. As a result, it is characterized by a long list of parameters that can be modified to increase performance, giving it excellent backup throughput. More importantly, NetVault demonstrates equally outstanding performance during restore operations.

During our backup tests, peak compression rates on individual files reached 2.96:1. That translates into an effective throughput of 88 MB per second of the Certance CL 400 drive. Taking the geometric mean of throughput over a series of backup jobs, results for the two drives were totally consistent with the results of our synthetic benchmark. Backup throughput on the Certance CL 400 was pegged at 51.7 MB per second. That level of performance was 5% faster than the throughput measured on the HP Ultrium 460, which was pegged at 49.2 MB per second.

The Certance LTO-2 drive demonstrated an slight but distinctly measurable edge in performance tests. Far more interesting is the SmartVerify scheme, which unfortunately cannot be readily measured or tested. Much like an insurance policy, it is the kind of thing that you don’t need until you need it. On the other hand, the inclusion of a fully licensed copy of NetVault provides a distinct value-add for small-business users who may not have an enterprise-class backup program in place.