Three Tape Libraries for Windows NT
Test Track: A review by ENT and Client/Server Labs looks at three tape libraries
Once the province of only the largest corporate networks, tape libraries have come down in recent years in both price and scale to the point that there is now a reasonable fit for almost any size company. Designed to automate the storage and handling of large numbers of tape cartridges and multiple tape drives, these devices remove the need for an administrator to monitor every operation of the backup or restoration process and to constantly switch individual tape cartridges into and out of the tape drives.
Such units are obviously useful for handling backup of information resources, one of the most critical things any data processing operation can do. Less obvious is the fact that a well-thought-out tape library may also make it practical to keep information available in "near-line" form for relatively easy retrieval.
Our testing revealed that these devices are becoming both mature and comparatively easy to use. The greatest distinguishing factor was the price, with the StorageTek unit being the most costly at slightly more than $56,000 for the hardware we tested. Pricing for the ATL units is set by the distributors, which we checked by consulting an Atlanta area reseller. The reseller also quoted a price near $56,000, but this was for a bundle that includes the library, a supply of data and cleaning tapes, installation and 1 year of service. Exabyte quotes a price of "slightly under $20,000" for the unit we tested.
In general, the units we tested posed very little trouble to install and set up. Indeed, the actual minute-to-minute operation of the products achieved the very desirable status of being "boring" for our testers. This coincides with our philosophy that flash-and-dazzle products are better left to end users: We want plain and simple stuff that works in a network infrastructure.
ENT and Client/Server Labs focused our examination on tape libraries that use Quantum Corp.'s popular DLT-7000 tape drive. Additionally, we limited the field to libraries supporting from two to four tape drives and 30 tape cartridges. In this group, we tested units from three manufacturers: the ATL P1000 from ATL Products, the Model 230D from Exabyte and the Model 9730 from StorageTek.
ATL Products Inc.
With its P1000 library, ATL has taken a decidedly different approach to the logical construction of a tape library device. Unlike its competitors, the P1000 is based on a PCI communications bus, like that found in the vast majority of new PCs. While the use of this technology for such a device is very new, it opens the possibility for some very interesting future expansions. Beyond simple SCSI connections to a host computer, the P-series libraries are designed to accommodate developments such as connection via Fiber Distributed Data Interchange (FDDI), or the inclusion of backup server processing directly in the library itself, as PCI adapters capable of supplying those services are developed.
The installation of the ATL P1000 was simple. One pleasant distinguishing characteristic of the configuration we tested was the use of very low-profile SCSI cables for connecting the components together at the back of the chassis. These connector cables were supplied as part of a thorough supply kit that included almost everything needed, down to cleaning swabs and lubricant for later service. One might hope, though, that the supply kit would be long forgotten by the time that operator service such as lubrication is required.
The internal physical arrangement of the ATL P1000 is also significantly different from that of its two competitors. Nine fixed tape storage slots are arranged in three rows at the back of the unit, with four DLT-7000 tape drives in two rows below that. At the front of the library are five more fixed storage bins, a tape loading port and two removable packs that hold eight tapes each, bringing the total capacity to 30 tapes. The tape-handling armature is located in the middle of the unit, between the tape storage areas at the front and back.
This arrangement has two consequences, the most notable of which was the absence of a viewport in the unit we tested. Because the front span of the library is fully populated with tape storage slots, no view is possible from the front panel. Our rack-mount test unit was not equipped with the viewport that would be found at the top of the floor-standing pedestal model. ATL explained that the port is omitted on the rack model because visibility through the top surface of rack-mounted equipment is limited. However, including the viewport would seem to present no obvious difficulty, while at least allowing the option of installing drives in a viewable location.
The other consequence of the internal structure is that the tape-handling mechanism is required to move in three axes: up and down, side to side and rotationally from front to back. In our test unit, the amount of sound and vibration produced by the library during its normal operation was noticeably greater than that of either competitor. This was especially apparent during the minute-and-a-half audit operation, when the library was checking the bar code and location of each tape. From the audible movement of the tape handler, our testers inferred that each tape was being moved out of and back into its storage slot, as well as being scanned. However, the lack of a viewing port meant that we were unable to verify that inference.
In the design of the operator panel, ATL has moved to the opposite extreme from StorageTek, providing a large, touch display and a rich set of functions with good graphical organization. An operator working through the front panel can control almost any aspect of the library behavior, from simply checking the content of a single storage location, through moving tapes between any two locations, and on to such detailed tasks as component calibration. Access to those functions is controlled through a two-level security code system, which establishes a four-digit numeric code for a library operator and a separate code for service access. However, owners of these units should not rely solely on access codes for security control. There is a backdoor security code, which is the same for all P1000 libraries.
Backup operations required much less tape and handler movement than the audit function had entailed, so the noise and vibration effects we noted in the startup actions may have little effect in real-world usage. The ATL library completed our test backup of 5.9 GB in approximately 25 1/2 minutes, somewhat faster than the competing devices.
However, ATL also gave us our only failure in normal operations. With each library, we used Seagate Backup Exec. with the AutoLoader feature from Seagate Software (Scotts Valley, Calif., www.smg.seagate.com) on our host system to conduct an inventory of tape contents, in which each tape is loaded to a tape drive and read to determine what if anything is recorded. On two occasions, the ATL unit failed to unload a tape from the drive when it had been read. In both instances, though, we were able to clear the problem using the front-panel operator interface.
Overall, we found the ATL P1000 library to be a good unit, with excellent operator controls. If it lacks any major feature, it is the physical visibility into the machine’s operations provided by the other two units.
In most of the feature categories we examined, the Exabyte 230D tape library sits squarely between its competitors. The operator interface is more full-featured than that of the StorageTek equipment, while the tape-handling mechanism is simpler than that of the ATL library. The 230D’s most distinguishing feature is its tape storage system, which aptly addresses the problem of loading and unloading tapes in sets.
Installation of our floor-standing test unit required only a screwdriver for tightening cables, though additional common tools would have been needed had we been using the rack-mounted version. Loading tapes took just a few minutes, though it required a couple of attempts to get the proper feel for seating the carriers involved.
Internally, the 230D stores tapes in removable data-cartridge magazines that mount onto a rotating spindle at the rear of the unit. Each of the six magazines holds five tapes, for a total capacity of 30 tapes. The two DLT-7000 drives are mounted below the storage spindle, along with a single, fixed slot for a cleaning cartridge. The handling armature is located to the front of the storage spindle and the tape drives, and moves in two axes: up and down or side-to-side. A laser bar-code scanner is attached to the handling armature, moving with it to read bar-coded labels attached to each tape. Though quieter than the mechanism in the ATL unit, the Exabyte library still operated with a noticeable level of noise and vibration.
The operator console on the front panel consisted of a moderately sized LCD panel and a set of eight programmable buttons below the LCD. The display was crisp and readable, with a display at the bottom showing what each button was configured to do at any given moment. Our testers found the branching menus fairly easy to follow, though on a few occasions an accidental extra key-press resulted in some confusion. A large viewport in the locking front door enables the operator to watch tape operations as they occur.
The most obvious difference between the Exabyte 230D and the other libraries is the lack of a cartridge access port. While this seemed to be a fairly significant shortcoming at first, that view changed as we came to understand the tape management philosophy behind the design. Unlike the ATL and StorageTek devices, the Exabyte 230D is designed for use in environments where tapes are handled in discrete sets, and those tape sets are moved between the library device and some separate storage location on a regular basis. In that light, the use of the five-tape carriers and the reliance on opening the unit to add or remove tapes makes considerable sense. The only confounding factor this design introduces is the need for more than ordinary care in controlling access to the locking key.
In startup and management operations, auditing the contents of the library took approximately 1 minute, placing the Exabyte library directly between its competitors for this operation. During the process, the unit scanned each of the tape labels to determine the identity and location of each tape. In backup operations, however, there was a noticeable difference in the time required for the 230D to copy our test data set. While each of the other units completed the backup in 25 to 28 minutes, the Exabyte required 41 minutes to perform the same operation.
Tests with several different tapes in both drives in the library produced the same results. An informal test of another DLT-7000 tape drive with a single-ended SCSI bus like that of the Exabyte produced results comparable to those of the ATL and StorageTek devices.
Several calls to Exabyte technical support failed to immediately resolve the problem. An Exabyte engineer visited our lab, and was able to determine that one of the tape drives in the library had an outdated version of the firmware installed. We installed a replacement drive, after which our backup procedures completed in 26 and 1/2 minutes, squarely in the range of the competitive products. Exabyte informed us that the correction could also have been handled as a software flash upgrade, had we chosen to do so.
In general, our testers liked the Exabyte 230D for its ease of installation and for its cartridge-oriented design. Tape handling, both by the unit itself and in manual operations, was easy to perform.
Storage Technology Corp. (StorageTek)
The StorageTek 9730 tape library is distinct from its competitors in several ways. Both the tape-handling mechanism and bar-code scanning hardware are more streamlined than the competing products. But the most immediately obvious difference is that the unit arrives in what might be described as kit form, with the DLT-7000 drives packaged separately from the library itself.
An excellent installation manual walked us carefully through almost every step in setting up and configuring the library. The library is capable of supporting from one to four DLT-7000 tape drives. Our test unit was supplied with the full complement. However, installing all four drives reduces the number of available tape storage slots in the library from 30 to 28. This is because a carrier with two storage slots must be removed from the system in order to accommodate the fourth drive.
Installing and connecting the drives was easy, requiring only a few standard tools likely to be found in any technician’s kit. An initial load of tapes is done by hand through the key-locked front door. A few of the slots were tightly placed for our more ham-handed testers, but presented no lasting obstacles.
Because installation was so easy and the instructions so clear and detailed (they go so far as to tell you when to open a door and how tightly to torque various screws), we were quite surprised when the system failed to start properly on the first attempt. Instead, the library halted during the process of auditing the number and location of tapes in the library, and presented an error code on its front panel. We eventually found the error description in a file on the diskette that accompanied the unit, and discovered that the library was looking for those two missing tape slots replaced by the fourth drive. The significant step of configuring the unit for the number of installed drives had been omitted from the otherwise exhaustive installation instructions. Completing that configuration cured the problem, and the system then started normally.
Internally, the tape-handling mechanism of the 9730 operates in two axes of motion: moving either up and down or turning on a central spindle. The three columns of tape slots and one of tape drives are arranged in an arc around the handler, like a miniature version of StorageTek’s larger, industrial-scale units. The bar-code assembly attached to the handler combines a row of LEDs for illumination with a charge-coupled device (CDC) for reading. The design of the handler and reader results in a significant reduction of moving parts compared with the other products, which StorageTek claims results in higher reliability. While we did not stress-test the mechanisms to failure, we certainly noted that the 9730 operated with much less noise and vibration than either of the other units.
StorageTek has also taken a minimalist approach with the design of its front-panel controls, providing only a small character-based display and four navigation buttons. A single-access port allows an operator to insert or remove a tape without opening the front door. Our testers found the menus relatively easy to navigate but were disappointed at the limited number of operations they could perform. For example, there did not seem to be any way from the control panel to move a tape between the access port and a storage slot or drive.
This limitation also became important when a power failure occurred during one of our tests. Not long enough to cause a shutdown of either the host computer or the library, it did disrupt the library and leave a tape in the handler mechanism. Shutting the system down fully and restarting it did not clear the mechanism. Correcting the problem required us to open the unit and retrieve the tape by hand.
During backup operations, the 9730 operated with almost total silence, moving tape cartridges between the storage cells and each tape drive with barely a whisper. Tape audit operations were also very fast and quiet, completing in an average of about 30 seconds for the full complement of 28 tapes we used. Backing up our test data set took approximately 28 minutes, comparable to the results with the other products.
Overall, we found the StorageTek 9730 a well-designed unit. It seems well suited to environments where it will be loaded with tapes, tasked by the host software and rarely touched. However, for those few times when operator-level service is required, StorageTek should consider making some significant enhancements to the available control options. A printed manual version of the error list is also called for.
Through the Test Track
To exercise the tape libraries in our comparison, we connected each unit to a Dell 4100/200 system with 256 KB of memory and dual Pentium 200-MHz processors. SCSI connection was through an Adaptec AIC-7880 SCSI host adapter. The ATL and StorageTek libraries we tested were equipped with differential SCSI buses, while the Exabyte library was equipped with a single-ended bus. Because the host adapter we used was a single-ended SCSI adapter, the differential bus equipment was connected through a differential to a single-ended converter.
The host system ran Windows NT Server 4.0 with Service Pack 3. A large SQL Server database was stored on the server, though SQL Server itself was halted during our testing, to simplify the environment. We used Seagate Backup Exec V7 from Seagate Software (Scotts Valley, Calif., www.smg.seagate.com), with the optional Autoloader feature to perform tape library operations. The entire content of the server (a data set of slightly less than 6 GB) was used for testing backup operations.