StorageWeek - The resource for Storage, Networking and Network Storage
www.storageweek.com
CBL Data Recovery
For Ad Information, call:
(818)STORAGE
(818)786-7243
NovaStor Ad
CBL Technology - Data Recovery
Link to First Looks
Link to Storage Reviews
Link to Software Reviews
Link to Backup Hardware Reviews
For Ad Information, call:
(818)STORAGE
(818)786-7243

Link to Benchmark Storage Innovations
 
Link to StorageWeek Home
Link to News
Link to Reviews
Link to Features
Link to About StorageWeek
Link to StorageWeek Labs
 
San Conference 2002
Link to Computer Associates - BrightStor
Link to Medusa Labs
Configuration Setup The Tests Results Conclusion Scorecard Bottom

 

Quantum|ATL M1500 Tape Library

By Mark Brownstein
Editor-In-Chief


       The Quantum|ATL M1500

Introduction

Quantum|ATL is probably best known for its large automated tape libraries (hence the name, ATL), ranging in capacities from the tens to the hundreds of tape cartridges.  (Note: In late October, 2001, Quantum|ATL was officially renamed the Quantum Data Protection Division --- ed) The M1500 is something of a departure from the rest of the product line. 

The M1500 defines a new category of tape library, providing expandability that allows the M1500 system to grow to match the increasing backup and restore needs of its users.  The basic M1500 is a 4U-high (7”) rack-width unit that is also available in a desktop version.  The original M1500 was designed to support one or two DLT8000 drives.  A robot in the center of the M1500 handles the task of moving cartridges between the tape magazines, located at the left and right sides of the library, and the drive(s), located near the back of the library.  The robot also moves the cartridges to a slot at the back. A second robot, designed for moving vertically, can take a cartridge and deliver it to a second M1500. As many as ten M1500s can be stacked in a single library.The interlibrary mechanism is called StackLink

Each magazine is easily removed.  The DLT version reviewed by StorageWeek held 10 cartridges per magazine.  An Import/Export slot brings the total tape capacity for a single library to 21 cartridges. With a native capacity of 40GB per cartridge, a single M1500 library has a native capacity of 800 GB.  Depending upon the data being backed up, each M1500 has a capacity ranging from less than 800 GB (for data sets that include a high proportion of uncompressible, or previously compressed files) to more than 1.6 TB (for files that are highly compressible). 

Quantum|ATL has also released a version that supports one or two SDLT drives. The native capacity of these drives is 110 GB, giving a native capacity of 2.2 TB, and a compressed capacity of 4.4 TB or more, depending again, on the data.  StorageWeek hopes to be able to review an SDLT version of the M1500 in the near future.

Additionally, Quantum|ATL has begun shipping a version of the M1500 that uses Ultrium LTO drives. As with SDLT, the capacity of each tape is 100 GB.  However, each tape magazine holds 12 tape cartridges.  With the Import/Export slot, a total of 25 tapes can be stored in each M1500.  The Import/Export slot is primarily designed for storage of a cleaning cartridge, in both the DLT and Ultrium versions, so it is not counted in total capacity estimates.  Native capacities for the Ultrium version of the M1500 are 2.4 TB, with compressed capacity of 4.8 TB, depending on the data being compressed. 


The M1500 can be stacked as many as 10 units high.  Here, it’s shown in a two library stack


As few as one M1500 library, as was reviewed here, or as many as 10 libraries can be configured into a single rack. Libraries can be added as needed.  When fully loaded, a library based on ten M1500s can provide the following capacities:

Drive Type
Max. Configuration Max. cartridges Native Capacity Compressed capacity
DLT8000 10 Libraries 200 cartridges 8 TB native 16 TB
SDLT 10 Libraries 200 cartridges 22 TB native 44 TB
Ultrium 10 10 Libraries 240 cartridges 24 TB native 48 TB

The M1500 is a solidly built library.  Its 75 pound body makes it more than a handful for a single installer – the company suggests that two people handle the installation.  It’s good advice. It’s a solid product. 

The M1500 is currently available in SCSI LVD or HVD versions.  The drives can occupy a different SCSI channel from the one used for the library automation.  Additionally, each library can also be attached to its own SCSI channel.  In theory, a backup server with multi-channel SCSI adapter cards can provide maximum performance for backup to multiple drives running concurrently. Having the ability to parcel out backup tasks to as many as 20 drives in the library system may be something of an advantage over larger libraries with fewer drives, or without the ability to address as many independent SCSI channels. The ability to grow with the user’s needs is a key point addressed with the M1500.

Top Introduction The Tests Results Conclusion Scorecard Bottom

 

Configuration

StorageWeek received an LVD version of the M1500, equipped with two Quantum DLT 8000 drives. The M1500 was connected using a single cable.  An active terminator was connected to the M1500.  This was the easy part.

We determined that at least one software package we used to test the M1500 was not able to write data to two drives at once, making our dual drive system somewhat more capable than necessary.  Another product, Backup Exec, supported the use of two drives simultaneously – we ran two different backup jobs simultaneously. 

We tried a number of backup software products with the M1500. Dantz Retrospect Backup 5.5 recognized the M1500 robot (as an M4 Data library), and the two DLT drives.  (It should be noted that Quantum|ATL acquired M4 Data, and took over development and marketing of the M4 / M1500 library).  Out of the box, Dantz Retrospect Backup supported the library.  Retrospect, and some other software products, were already able to support the M4 Library, which had been available for about a year.

Veritas BackupExec 8.6 was also installed during the review.  As with Retrospect, it provided automatic support for the library.

Control of the library robotics by both products was flawless.  The drives loaded and formatted tape cartridges as expected. 

In order to get the best measurable performance, we removed most of the compressed files, plus the temporary Internet files, from the list of files to be backed up – filter actions that should have minimized the impact of non-compressible files.  Typically, compressed files actually increase the amount of time taken to perform file backups on drives with hardware compression enabled (as was the case with the DLT8000 drives in the M1500). 

Although we wanted to be able to test performance in both native (without hardware compression) and compressed modes, the method for toggling hardware compression on and off was not easy to determine.  In most installations, hardware compression, a long-proven technology that is standard on most tape drives, is usually always left enabled.

Setup

After testing was completed using Retrospect Backup, Retrospect was disabled.  When Retrospect was originally installed, it was configured so that it did not load a backup agent when Windows is launched. 

Backup Exec Version 8.6 was installed, and configured to run test backups.  One nice feature on 8.6 was its ability to write simultaneously to both drives.  In the case of StorageWeek’s testing, we performed a full backup of the server with one DLT8000 drive backing up the server’s C: drive (with about 60 GB of files), and the second DLT8000 simultaneously backing up the remaining physical and logical drives, again totaling approximately 60 GB of data.

Conceivably, the use of more than one drive would also be very useful for such tasks as, say, backing up the data server on one drive, and backing up network drives on the second.  A system with more than one M1500 could be used to backup many drives on a network, with one or more clients assigned to its own drive, and multiple simultaneous sessions running.  If each library used a dedicated SCSI channel, considerable backup performance is expected to be possible.

Controls

An LCD display on the front of the M1500 provided instant status information, and also provided the interface to a rather complicated options menu.  Controls consisted of six buttons: at the outer edges of the control area, buttons enabled the release of tape magazines. A lock is located to the right of the left magazine release button.  Two rectangular buttons – one labeled MENU/ESC and one labeled ENTER flank an up arrow and a down arrow.  The combination of the four buttons – MENU/ESC, up arrow, down arrow, and ENTER are used for all front control operations. 

The LCD display panel, located above the control keys, provides the basic user interface.  During start-up, it reports on the various steps performed to calibrate the robot, inventory the magazines, and calibrate the drives.  When accessed using the MENU/ESC key, the library is taken off line, enabling a wide range of library settings to be implemented (but, regrettably, no control over the status of hardware compression).  The settings allow a user to change SCSI ID letters for the drives and the library, control parity and other SCSI settings, run automated tests (including robotic exercises), and perform other configuration and monitoring tests.  Additionally, usage information and control of media movement can also be handled through this menu.  Further, the insertion and removal of a tape for storage in the Import/Export slot can also be managed through this menu.

A secondary menu, the Quick Menu, can be accessed by pressing the ENTER button.  This enables the user to perform a basic inventory of the drives and library.  It also allows the user to turn power to the drives on or off.  A valuable feature is that these tasks can be performed without taking the library off line. 

Firmware upgrades and diagnostics can be performed by connecting to a serial port at the back of the library.  During the course of this review, we upgraded the Library firmware with no problems.

Top Introduction Configuration Results Conclusion Scorecard Bottom

 

The Tests

It was our plan to establish a fixed data set, consisting of drives on the backup server, and drives on two or three clients, for the testing.  The plan included backup of a RAID array, as well as backup of individual drives, and backup of drives in an external SCSI enclosure.  .

The goal was to be able to compare apples to apples, running tests against the same data, from the same drives, regardless of which software was being tested; and regardless of the drive or library being tested.  Although this was the ideal, it didn’t work exactly this way.

Early testing included one 60 GB IDE drive, two 18 GB SCSI hard drives mounted internally, two or three 60 GB IDE drives, and four SCSI hard drives (one 9 GB drive, one 18 GB drive with a single 9 GB partition, one 18 GB drive with an 18 GB partition, and one 47 GB drive, with approximately 35 GB of data) in an external enclosure.  The two 60 GB IDE drives were alternately equipped as two standalone drives, or as parts of a 120 GB IDE RAID5 array. 

Testing was performed on this full system, with about 110 GB of total data (minus the compressed files that we excluded in order to boost performance), in our early tests.  Although we encountered failure of more than one hard drive over the course of the review, the data was restored from tapes recorded using the M1500, onto new drives.  Additionally, we removed the external SCSI enclosure from our test platform, moving the data on its drives to other drives in the test system.

We still had more than 110 GB of data to be backed up and restored, although it was on a significantly different set of drives.  Additionally, we frequently defragmented the individual hard drives before performing backup, in order to reduce delays in transfer speeds that could have resulted from drive fragmentation. 

Results

I agonized over getting the best possible performance from the drives in the M1500.  I tried different SCSI controllers.  I tried different SCSI drivers.  I got varying results.

I tried backing up drive arrays, assuming that I would get better performance from drives that had striped the data, and that were able to use multiple actuators to read the data, maintaining higher speed data transfer.  I tried backing up with other processes running on the system.  I tried backing up with the bare minimum of tasks running on the system. 

I tried a variety of software applications.  I tried different SCSI cables for connecting to the M1500.

The issues I had encountered were related to getting the maximum performance out of the tape drives, and were unrelated to the library or its robotics.  The DLT8000 drives in the M1500 we tested are basically the same as the DLT8000 standalone drives in use at many IT shops, and the same as those in other libraries available from Quantum|ATL and others.

Issues I faced during the course of the review are not further discussed in this review.  The issues can be found in a feature article on the StorageWeek site.  These issues are related to high performance SCSI tape drives, and don’t reflect negatively on the M1500.

The fact that the M1500 now comes with SDLT or with Ultrium drives reinforces the basic fact that the important differentiators between libraries are the flexibility, capacity, user interface and management and, to a limited extent, media management speeds.  Once you get past the robotics, most of the performance issues are related to the performance of the drives.  Seating a part time employee next to the drive, and waiting for the employee to swap out cartridges each time one fills up will probably provide performance figures similar to those of a tape library (if the employee doesn’t fall asleep). 

It is important that a library is able to accurately maintain an inventory of cartridges in the library.  The ability to read a bar code is an important feature that the M1500 supported.  For media changer systems with only a small number of tapes, tape inventory is not a particularly important feature, but when considering a library that begins with the ability to manage 20 cartridges (24 in the case of the Ultrium version) – and can be expanded to as many as 240 cartridges (in the Ultrium version), being able to track and identify each cartridge becomes much more significant.  Although a bar code reader is an option available on some other sub-20 cartridge devices, the reader is included as a standard feature in the M1500.

The M1500 performed all tape management tasks flawlessly.  It was able to quickly move tape from one slot to another smoothly and accurately.  It was always able to grab cartridges when ejected from the drive, and was always able to load tapes into the drives without jams or other errors.  The front panel provided a real-time display of the current tasks being performed and status of the library.

We were unable to test integration with other M1500 libraries, so tests on transfers between libraries couldn’t be performed.  If we’re fortunate enough to add a second M1500 to the test setup, this would enable additional tests.

Reliability

The M1500 was powered on, and frequently used, during more than four months of testing.  The two DLT8000 drives are drives from a family with a long record for reliability.  Because they are hot swappable, in the event of a drive failure, the failed drive can be easily swapped out.  The performance history for DLT8000 drives has been good.

Until survival data becomes available for the LTO Ultrium and SDLT drives that are available in other versions of the M1500, some caution may be prudent before a wide scale adoption of these devices is made.  However, we are unaware of reports of reliability problems with either of these new drives. 

In general, library robotics have been developed to a point where, overall, they are very reliable.  It is expected that the robot in the M1500 will retain its reliability for many years.

Top Introduction Configuration The Tests Results Scorecard Bottom
Conclusion

The M1500 is an interesting device.  When it was first presented for review, it was not yet available as a Quantum|ATL product, and software support was limited.  Since that time, additional software support has become available.  Additionally, SDLT drives have become available for the M1500, and an Ultrium version was announced. 

Robotics operated quickly and accurately.  A quick inventory of the library, in which the robot scanned the bar codes of the media in each magazine was rapid and accurate. The process of doing a full inventory of the library, which involved using the drive to read data from each cartridge, was limited by the speeds with which the drives could calibrate and read the new media. 

Its predominantly black face, with judicious use of blue on the magazine door latches and around the control panel provide an attractive, industrial look. (Quantum has changed the cosmetics of the M1500, but still delivers an attractive unit with a definite industrial appearance - ed.) A clear window, through which you can see the robotics, drives and magazines, all bathed in a red light, can provide for endless entertainment when the TV goes on the blink.  It’s a nice looking device.

The M1500 is a solid product – it was able to run, virtually non-stop, for many months.  Its built-in redundancy, supporting hot swap and hot plug for disk drives, fans and power supply qualify it as a true Enterprise level backup product.  The ability to failover an M1500 library to a reserve library in the same rack further underscores this product’s suitability for use in a growing Enterprise or department. 

The expandability of the system, allowing it to grow to become a library that can handle 200 cartridges (or 240, in the case of the Ultrium version), provide a considerable amount of growth potential that is available in few other products (Overland Data Systems' Neo is also expandable).  In the case of the unit we reviewed, equipped with DLT8000 drives, the ability to replace the drives with SDLT drives at some time in the future may also be a useful feature for companies that don’t have the budget for SDLT drives, but requiring solid backup with a tangible growth path.  Being able to make the future upgrade, without having to purchase a new library system, should result in significant future cost savings for such a company.

It should be noted here that the Ultrium version is somewhat different from the DLT version.  It uses magazines that are different from the DLT version, and is not able to share media with DLT. This should be obvious. 

The M1500 is available in a rack mount version, and in a desktop version. For expansion, the desktop version can be easily modified to be used in a rack.

The ability to control each drive on a separate SCSI channel, if desired, can assure the highest performance possible for backup.  Although the current LVD and HVD support should be more than adequate for drives many years into the future, the M1500 may also be upgradable to future interface standards (can fibre channel and iSCSI be more than a year or two away?), because the M1500 uses a standard PCI connector to provide the data interface.  Quantum|ATL has not announced plans for other interfaces, but the potential for support of new standards has been built into the M1500.

The M1500 has earned our solid recommendation for companies looking for an industrial strength backup library.  It gets an even stronger recommendation if expandability – both in the number of drives and cartridges that can be supported, and for adding or upgrading drives—are major concerns for an Enterprise looking for a long-term backup system.

StorageWeek Scorecard

Setup

Features

Performance

Serviceability

Value

Overall

3.75

4.0

4.0

4.0

4.25

4.0

Category scores are awarded on a scale of 1 to 5 (5 being best). In general, few products will earn a 1 or a 5.  Key: 1.  Features are missing or product is lacking value in this category. 2.  The product may meet basic criteria, but is marginally acceptable.  3.  Product meets basic requirements, but has little to make it stand out in its category.  4.  Product is better than average.  Product has better than expected capabilities, or is judged to have capabilities or meet requirements that make this considerably better than average.  5.  Excellent product, far exceeding requirements and expectations.  A 5 is, in most cases, unattainable.


HOW WE TESTED

Our backup server was a custom built system based on a 1 GHz Athlon Thunderbird CPU, with 512 MB of RAM.  The system was running Windows 2000 Server.  The hard drives were changed during testing.  The server had one 60 GB (later upgraded to 80 GB) IDE hard drive, which was used as the boot drive, and to support the backup software applications.  Additionally, a 160 GB striped IDE RAID Array, based on the HighPoint HotRod 100 Pro chip was implemented inside the backup server.  The array was based on two 80 GB drives, striped into a RAID5 array.

A third IDE drive was also connected to the RAID controller, but as an independent drive.  This drive was changed during testing.  Initially, a 7200 RPM 60 GB drive was used in this array.  Ultimately, an 80 GB drive was used.

Internally, two 18 GB UltraSCSI 160 drives were used.  Initially, four SCSI drives, installed in an external Kingston DataSilo were used.  The capacities of these drives were 18, 18, 18 and 47 GB.  Problems with these drives ultimately required that their data be copied to the IDE array, or copied from backup tapes to the array.  The DataSilo was removed from the system.

A 1 GHz Intel NIC provided the connection to network clients.  Network clients included a Compaq ML370 server with two 9 GB SCSI hard drives, a Compaq Desqpro EN with one 9 GB hard drive, and a custom built system with three IDE drives. 

Testing involved a combination of backup strategies.  These included backup of each individual drive in the backup server, and a complete backup of the drives directly attached to the server.  The backups also involved backup of the drives in each client, and a backup of all drives on the network. 

Backup also featured full backups – including temporary Internet files, in addition to previously compressed files.  Additionally backups of complete sets, with most of the previously compressed files and temporary Internet files excluded were also performed. 

The majority of tests were performed with Dantz Retrospect 5.5.  Among the reasons that this product was used were its excellent support for the M1500 library, and its useful, frequently updated progress reporting.  Additionally, a log provided a useful list of all operations. 

We also tested with Veritas Backup Exec 8.6.  Although the status reports were not as informative, the ability to support multiple simultaneous processes (such as the creation of one backup set on each of the two drives) was also a very useful feature.

Data sets ranging in size from less than 3 GB, up to more than 120 GB were written to the M1500, using both software products.  Data was restored, as needed, from backup sets created using Dantz Retrospect. 

Vendor Contact Information:

Quantum Corporation
Data Protection Division
141 Innovation Drive
Irvine, CA 92612-3040

Phone: (949) 856-7800
Fax:      (949) 856-7799

www.QuantumATL.com

©2001-2006

StorageWeek and Mark Brownstein

 
Copyright 2002 StorageWeek
Link to StorageWorld Conference
News Reviews Features About StorageWeek StorageWeek Labs