The saying, "an ounce of prevention is worth a pound of cure," is never truer than when it applies to backing up your data. We look at five of the best tape drives to help you decide which one is right for your Linux system.
From the most hi-tech dot-com to the oldest brick-and-mortar company, any business today is only as good as the information stored in its computers. As a result, conducting a regular backup is one of the most important chores for any system administrator. The increasing use of central storage (even for PC-based networks) simplifies the task to some extent. However, dependence on servers with ever-growing disk drive capacities necessitates the growth of back-up systems.
This article surveys some of the back-up tape drive systems available for Linux servers and workstations today. Since there are several different products available, we decided to take a look at tape drives intended for the smaller servers found in small businesses and fair-sized departments (the primary market for Linux servers). Our goal was to back up at least 10 GB of data without manual intervention (so that back-up could be a “fire and for-get” operation). We wanted each back-up session to use a single piece of media.
We tested five tape drives from four manufacturers: the Ecrix VXA-1, the Quantum DLT 8000, the Hewlett-Packard SureStore DAT40e, and the Seagate Tapestor Travan 20 and Tapestor Travan 20NS.
None of these tape drives are packaged specifically for Linux, which means that we had to find our own back-up software to drive the units. So, we downloaded the Arkeia package from Knox Software (http://www.arkeia.com) and found that, in all but one instance, every drive we had was supported directly by Knox’s package. The exception was the OnTrack ADR-50 drive, for which support is supposed to be added in December. We ran the Arkeia package on a system loaded with the Red Hat 6.2 distribution.
To accurately simulate a server that is still online during a back-up session, we selected a Pentium 166 computer with 40 MB of RAM as our test vehicle. Our intent was to simulate the performance of a tape drive run on a server under load; we attempted to do this by limiting the processing power that was available. We used the exact same test bed when evaluating every drive so that comparative times would remain valid.
For the Seagate Travan drives, we used a $65 Adaptec 2906 SCSI adapter; for the remainder of the drives we used an Adaptec AAA-131U2 SCSI RAID adapter. While the RAID has a SCSI port acceptable to the Travan drives, we were interested in how this particular drive would work with a low-cost SCSI card.
Installing each product was quite easy with Red Hat’s automatic hardware recognition feature. The Arkeia software was configured for each supported tape drive with very little difficulty. Expect the installation of the back-up software and tape drive to take no more than three hours for internal tape drives and two for external ones.
Seagate Tapestor Travan 20
Seagate Storage Products
1650 Sunflower Avenue
Costa Mesa, CA 92626
(800) SEAGATE [732-4283]
For small organizations or workgroups, particularly those with little money to spare, the Seagate Travan 20 is an excellent fit. For just under $800 you can purchase the drive, an inexpensive SCSI adapter (such as the Adaptec AHA-2906 we used), and enough media to run backups for several months.
If you want to take the computational load of compression off of the server and place it in the drive, you can fork out an additional $190 and get the Travan 20 NS drive. The Travan 20 NS looks and performs the same as the Travan 20, except that the drive’s electronics perform data compression on the fly. Offloading the processing improves back-up time by eliminating the need for the server to compress data before sending it to the drive.
The time required to perform a backup of 500 MB using these internal drives was just under 15 minutes for the Travan drive and 13.5 for the Travan NS drive, which is about the same as the other drives we tested. To restore our single file, the two drives required almost eight minutes — much longer than the other drives we looked at. The capacity of each cartridge is 10 MB minimum — up to 20 MB when you are backing up non-compressed files and using software compression with the Travan drive, or the built-in compression of the Travan NS drive.
The lower initial investment of the two Travan drives is offset by a slightly higher running cost (media is about $3.80 per GB if you use the entire tape) and slower individual-file restore times. Recovering a mistakenly deleted file will take longer with the Travan than with any of the other drives we tested. Additionally, there is currently no upgrade path to larger-capacity tapes, so when you outgrow the Travan you will need to either replace it or use multiple tapes per back-up session (meaning a sysadmin will need to be present).
These drives are installed in the same type of case opening as a floppy drive. You can purchase an external case if you wish; expect to pay $80 for such an item. Standard jumper blocks are used to set the SCSI ID and SCSI bus termination options. The interface is the standard 50-pin SCSI drive used on SCSI disk drives and similar devices. Many RAID cards, including the Adaptec series, include a 50-pin connector for just this purpose. The drive gets its power from the server. Installing these drives is as easy as installing a SCSI CD-ROM drive into your computer; simply set the SCSI address with jumpers, connect the cable to your SCSI controller, and bolt the drive into place.
Frankly, putting the drive into the server could be an advantage for small shops, particularly for those that do not employ a fulltime system administrator. You won’t have to worry about the drive being knocked off of a table or getting unplugged by accident. This drive is also an excellent choice for standalone workstations that need to back up large files, such as Linux-based CAD and scientific data-analysis stations.
Seagate Tapestor Travan 20
Capacity (non-compressed): 10 GB
Interface: SCSI 50-pin
(ATAPI also available)
Size: 3.5-inch internal
Media MSRP: $38/10 GB
5525 Central Avenue
Boulder, CO 80301
What fits in the space of a CD-ROM drive, uses cartridges the same size as those used for camcorders, and adjusts to the rate the computer can supply or accept data? The Ecrix VXA-1 drive stands apart from the other drives in this article by adapting to the system, rather than making the system adapt to it. And as provided, the drive can handle multiple-sized cartridges
of up to 33 GB. The built-in hardware compression doubles the capacity (assuming fluffy data) to 66 GB. If you prefer, the drive is also available as an external drive for $1,149.
Setting up this drive is exactly like setting up a CD-ROM drive for SCSI. Ecrix provides jumper blocks with little handles, so you don’t need tools to change the SCSI address from its factory default (even if you have fat fingers like I do). Plug in the Ultra2 SCSI cable (and terminator, if this is the last device) and the standard power, tighten the bay screws, and you are done. Both Red Hat 6.2 and the Arkeia back-up software (the software we used in this article) recognize the drive.
In our hands-on work with the drive, a backup of 500 MB of data required a total time of just over 14 minutes. Restoring our single file from the end of the tape required 85 seconds. These times fall in the middle of the range we experienced with the drives in this article. During the back-up process, the drive was running at part-speed; we expect a faster server would run the data out more quickly.
Unlike Exabyte drives and DAT drives, the VXA drive is not a streaming tape drive. Instead, it uses a different recording format, variable speed operation, multiple-trace reading, and read-after-write data checking.
The recording format changes ensure that defects in the tape and electronics have a minimal effect on the ability of the drive to accurately recover data. The variable speed operation works exactly like the half-inch magnetic tape drives of mainframe days; when the data stops, the drive stops. The use of overlapping read heads saves us the time-consuming recovery cycle of backing the tape up and retrying if it is misaligned. Instead, the blocks are short enough that one of the read heads will read the block of data in spite of misalignment. Finally, the use of read-after-write recording means that bad data can be re-recorded without backing up the tape; the read mechanism can then reassemble the fragmented block.
Data integrity is ensured using several error-detection and correction schemes, not unlike those used for DDS-DAT and CD-ROM recording.
The adaptive speed adjustment made itself evident when the software was required to read a large number of very small files. Unlike the other back-up drives, the Ecrix VXA drive was able to negotiate the slow data writes without reversing the tape.
The incremental cost of using the Ecrix drive is second only to the HP drive, at around $2.25 per GB, which makes this system attractive for larger server installations that don’t have a full-time sysadmin. The 33 GB tape capacity will make out-growing this system tough. If you aren’t cash-strapped, this is the system to get.
Capacity (non-compressed): 33 GB
Interface: SCSI Ultra2 LVD/SE
Size: internal, 5.25-inch half-height
Media MSRP: $30/12 GB, $45/20 GB, $80/33 GB
$2.50, $2.25, $2.43 per gigabyte
Quantum DLT 8000
500 McCarthy Blvd.
Milpitas, CA 95035
The Quantum DLT 8000 drive is large, a bit clunky, and decidedly the most expensive tape drive we looked at. The media it uses is half-inch magnetic tape, and unlike the cartridge used by every other drive, this cartridge doesn’t contain the tape all the time; the tape is pulled out for use and then rewound into the cartridge before it’s ejected. On the other hand, the technology used in this drive is the most mature, and the tape system is gentle enough that the tape is rated for over a million passes.
In terms of back-up performance, it holds its own, taking 16 minutes to back up a 500 MB file set. Retrieval of our single file required almost three minutes (longer than the Ecrix and HP DAT, but shorter than the Seagate Trevan drives). This is the largest capacity drive we looked at; each cartridge holds up to 40 GB of data before any compression. In theory, you could hold up to 80 GB of uncompressed information (say from a database) on a single cartridge.
For such a large cartridge, there is very little space for a label. One dodge would be to label both the cartridge and the plastic case with a barcode and then place any descriptive labels on the plastic case.
Unlike the other drives that let you just slip in a cartridge, the Quantum requires you to pull up on a handle before you can insert a tape cartridge; when you are finished, you have to let the drive suck the tape back before you can lift the handle. The instructions contain dire warnings illustrating the events that may transpire should the steps not be performed in that order. We suggest using a good UPS to protect both the server and the tape drive so you can unload the tape in the event of a power failure.
This expensive drive costs $4,700, and the media is $99 for a 40 GB cartridge. The cost per GB is around $2.50 (about the same as for the Ecrix drive). For our target small-server shop, this drive is overkill…but it’s nice to know that when your server backup needs to grow, you have this option. However, for a large shop, the cost of the drive is well worth it. For even larger shops, there are versions of this drive with magazines to handle multiple tapes at once. For really large shops (or if you are doing set decoration for the movie Eraser), the cartridges from the Quantum DLT 8000 can move to a StorageTek TimberWolf 9740 tape library (with rotating robot arm). But, that’s beyond the scope of this article.
Capacity (non-compressed): 40 GB
Interface: SCSI Ultra2 LVD/SE
Size: external, 3.38×8.93×11.25
Media MSRP: $99/40 GB
HP SureStore DAT40e
3000 Hanover Street
Palo Alto, CA 94304-1185
The Hewlett-Packard SureStore DAT40e is an external drive that is attractive for a site needing both back-up capability and data interchange from system to system and location to location. Indeed, the tiny size of the cartridge makes it ideal for companies that use services such as “FedExNet” data exchange and that have outgrown the use of optical media and CD-RW.
Once you have a suitable SCSI adapter and 68-pin Ultra2 SCSI cable, installation is a breeze. A switch in the back of the drive sets the SCSI address. The device can be inserted into a daisy-chain with ease, so you can have multiple disks and tape drives; just in case you want to use the DAT40e as the last drive in the chain, HP supplies a terminator. The drive requires a 4.65×9.00 inch space on the shelf, table, or rack; the drive is 4.05 inches high.
In our hands-on use of the drive with the Arkeia back-up software, the back-up process took just under 14 minutes. When we restored a single file though, it was clearly the fastest of those we reviewed — 42 seconds to seek and load the file back to the system. This is due to the indexing function built into DAT systems, designed originally to let a user choose an audio cut to play.
Hewlett-Packard claims the DAT40 drive can read all older-technology DAT tapes except the original 60-meter tapes.
HP’s list price of $1,365 makes the DAT40e a reasonably priced back-up system to purchase. Media cost is around $1.80 per GB if you use the entire tape. If you don’t already have a SCSI Ultra2 controller, such as those used in RAID systems, then you should expect to pay just under $300 for a suitable SCSI controller (such as the Adaptec 29160).
Digital Data Storage (DDS) technology has come a long way since its introduction in 1989. What started as an audiotape standard has grown into a viable data storage and interchange standard. Capacity has grown from the original 1.3 GB of data to 20 GB and will double again in the next 18 months. This capacity is based on data stored before compression; when the files being backed up are not already compressed, each cartridge handles 40 GB.
Of the five tape systems we looked at, the DDS-4 cartridges are clearly the smallest. On the plus side, this makes the tapes easy to store; on the minus side, the cartridge design severely limits the kinds of labels that can be placed onto it. This is particularly bad for sites that are in the habit of bar-coding tapes for identification purposes.
DAT has proven itself reliable in data back-up applications, and both drive and media improvements have increased the storage density. The small footprint of the drive, and of the media itself, makes the HP solution attractive for a medium-sized server. When your needs outgrow the basic drive, an autoloader version upgrade will expand capacity to 100 GB online.
Capacity (non-compressed): 20 GB
Interface: SCSI Ultra2 LVD/SE
Size: external, 4.65x9x4.05
Media MSRP: $36/20 GB
OnStream ADR50 — Coming Soon
The ADR50 drive is a 25 GB tape drive system that is undergoing changes to work with popular Linux back-up software like the Knox Software Arkeia package. We had the chance to look at an external drive, which lists for $949; an internal version is available for $799. The media is $50 for 25 GB, or $2 per GB, which puts it in the same ongoing price range as the slightly lower-capacity HP DDS/ DAT drive.
When we talked with Knox Software about why the ADR50 wasn’t supported, we were told that OnStream had made changes that would allow Knox to support the drive more easily, and that full support would be available quickly…but not quickly enough for the drive to be included in this article.
Like the Ecrix, OnStream supports variable-speed operation so the drive can adapt to the data rate supported by the connected computer. The 8-gap head reads and writes bytes in parallel, reducing the tape-to-head speed and still maintaining high throughput.
Connecting the external drive is easy: connect the Ultra2 SCSI cable and the power cord, and then boot Red Hat so it can recognize the drive. The cartridge is a little unusual, in that it is about the size of a thin paperback book — finding a media case to fit the cartridge should be interesting…
In any case, we look forward to testing this unit and reporting our findings in a future review.
Stephen Satchell is a freelance writer living in Nevada. He can be reached at firstname.lastname@example.org.