Tape libraries automate backup

Everything you need to know about tape libraries.

This Content Component encountered an error
This article can also be found in the Premium Editorial Download: Storage magazine: Is storage virtualization ready for the masses?:

Tape and recovery are synonymous. But it's the automation that libraries add to the base technology that make tape a viable road forward for storage managers.

The comparative low cost for backup and recovery ensures tape's continued use to keep up with the explosion of data. New tape technologies need to maintain pace with the increasing time and cost of backup/recovery. Even though redundant and/or remote disk mirroring has become less expensive, and thus more popular, due to its higher speed, tape is still the best solution for many businesses.

Best practices for recovery take a data snapshot from disk and write it to removable tape cartridges. Then the cartridges are sent to a safe, off-site archive. Tape overcomes single points of disk failure like software corruption cascading to the mirror or disasters, including sabotage, affecting both locations. For example, several companies mirrored from one World Trade Center tower to the other before Sept. 11. Others mirrored from their WTC tower to buildings across the street, which were also destroyed.

When using tape, a library is the most efficient method for fast, reliable backup and restoration of large amounts of data. Library automation provides economies of scale that individual drives can't by consolidating data, lowering total cost of ownership (TCO), reducing human intervention and error, simplifying backup and recovery, providing unattended lights-out backup and allowing a scalable growth path. Libraries' reduced administrative costs have a big impact on TCO. That effect is amplified by using a tape library in conjunction with a storage area network (SAN). But all tape libraries are not created equal, and choosing the right one requires extensive analysis (see "How to choose a tape library").

Tape library configuration
Before you can calculate a tape library's TCO, you need to choose the right combination of tape formats (see "A guide to tape formats") - which determines the capacity and transfer rate of individual drives - and the number of tape slots, which determines the total capacity of your library. The range available to you is quite broad: from autoloaders costing a couple of thousand dollars, to enterprise libraries in the hundreds of thousands of dollars (see "Scoping out tape library vendors").


How to choose a tape library

All tape libraries are not created equal, and all their features aren't equally important. Here's what to focus on.
PRIORITY KEY FEATURES HOW TO COMPARE
Very High Density Total capacity measured in GB/cu ft. = GB/U or ft. For automation use, the higher the density the better.
Throughput Tape drive's transfer rate x number of drives = throughput in GB/hr. Most important to meet recovery time objective, not just backup window. MB/s x 3.2 = GB/hr is ideal throughput take off. Allow 15% for overhead.
High Drive types Make sure your tape drives and popular tape drives are supported.
Maximum cartridge capacity More cartridges for more capacity. Each slot holds one tape cartridge.
Maximum drive capacity More drives for faster data transfer. Most start with two or four drives so you can add drives later, when you need faster transfer rates.
Maximum library capacity Tape cartridge capacity x total slots = total GB or TB.
Price After meeting above criteria, price is a key differentiator. Make sure and look at total cost of ownership (TCO).
Warranty Varies by vendor from one to three years and on-site to parts exchange.
Medium Remote monitor Allows monitoring library from any Web-based terminal.
FC attach Native Fibre Channel allows easy SAN attachment and reduces components.
Average cell to cell Speed of robotics affects access time to data, as does cartridge exchange and tape drive load time to data. Faster is better.
Cubic feet or U Physical volume or size of library.
Cartridge access slot Enter cartridges into library without opening door; some use carriers.
Low to
medium
Average power consumption Measured in watts. Correlates with ongoing operating cost. Larger libraries are 7-8 times more costly than smaller ones.
Low Mixed media Supports different drive and cartridge types in same library.
Cleaning slot Special location for cleaning tape so you don't have to use access slot.
Mean time between failure Expected reliability measure. All libraries have high ratings and require preventative maintenance.
Height Self explanatory, but make sure it fits through door and ceiling.
Weight of library Weight varies significantly, higher densities are sometimes lighter.
Source: Evaluator Group

The first step in automation is an autoloader. An autoloader, by definition, has one drive that typically serves seven or eight cartridges. An autoloader is good for unattended backup for a week or more for smaller companies. The newest versions are rack-mountable in a 2U format and have two drives - in case one fails. A library, in contrast, has more than one drive. Libraries range in size from small footprints up to room-size models comprised of chained-together libraries serving an entire enterprise. Libraries usually have express cartridge access to load an individual tape or groups of tapes. Larger libraries have bar code readers to inventory and keep track of tapes, magazine-style express slots and a host of other features.


Scoping out tape library vendors
COMPANY AUTOLOADER LIBRARY
FOR MIDRANGE
LARGE LIBRARY
FOR ENTERPRISE
ADIC
ATL
Benchmark NO NO
Exabyte
HP
IBM
Overland Data NO
Sony NO NO
StorageTek
Source: Evaluator Group; * Denotes 2U rackmount
A word of caution about large libraries that read bar-coded tapes: All bar code readers are not created equal. You may get errors reading bar codes when mixing different automation vendors. Worse yet, you may have errors reading the particular bar code already on your large library of tapes. Send the vendor some of your standard bar code labels. Have them test and confirm that their model can read your labels.

Library vendors keep increasing system densities to decrease footprint. Another trend in library design is the use of auxiliary memory on cartridges for faster operation and more intelligent diagnostics. Libraries typically have a life of five to seven years and the tape drive technology may be upgraded during this time. All major computing platform environments are supported.

Libraries allow two or more drives to access a large number of cartridges. The transfer rate for the library is essentially equal to the transfer rate for a single drive multiplied by the number of drives, minus a little overhead, which all adds up to transfer rates of over a terabyte of data per hour. New formats and migration roadmaps associated with tape formats like LTO, S-AIT and SDLT, along with faster robotics, are raising the bar on capacities and transfer rates. Libraries are denser, smaller and faster than anyone thought possible a few years ago.

When sizing automation to an application, consider the ratio of drives to cartridges. For capacity, fewer drives are used. For transfer speed, more drives are used. Make sure that you have the actual transfer rate (less overhead) needed to meet your business recovery time objective.

Use TCO analysis to evaluate libraries
Vendors traditionally sold libraries via relationships, strong product features and functions. Users looked at TCO as the cost of hardware, software, maintenance contracts, power use and personnel to install, operate, manage and maintain an IT system - the technology TCO. As technology costs continue to drop and personnel and business costs continue to rise, business issues have assumed more significance and can no longer be left out of the TCO assessment.

Libraries should be selected based on how well they support the business TCO. The business TCO addresses the potential dollar cost/benefit to the user and business, showing how well the proposed library improves availability, performance and recovery cost impacts. Business costs have the greatest impact on a company's recovery from an outage and must be part of the TCO evaluation (see http://www.evaluatorgroup.com for a free white paper detailing TCO calculation).

You won't want to send a library back once installed, so include an on-site service contract in your calculations. Make sure you get the service level you need. It's a good idea to have a spare tape drive on hand: Will vendors provide a spare drive? Media cost can be the largest expense in large systems: Is there a special media purchase price bundle with a new library? Tapes can be reused, but need to be retired when they have been used for specified passes, or begin to show a rise in soft errors reported by the backup software. Include that in your TCO.

Choosing a tape format
There are more choices today for tape drives than ever before. There is no perfect format for all of your applications - you should consider the best choices for work groups, midrange systems, enterprise systems and specialized applications on their own merits. Unfortunately, the interchange capability of different tape technologies is nonexistent or extremely limited. There is interchange between vendors of the same type of devices. For practical reasons and safety, make sure the interchange is demonstrated and certified.

Comparing the capacity of different formats can be tricky. Most vendors offer data compression for tape systems built into the hardware. Also, independent software vendors (ISVs) provide compression. A typical loss-less data compression rate to use for estimation is two to three times, although the algorithms vary and the compressibility of data varies widely. Generally, business data compresses well, scientific data compresses poorly and systems data varies. To compare different tape technologies, always use native, uncompressed specifications. But you should also consider how cartridge construction, recording technology, software support and other tape characteristics meet your needs.

Cartridges.Tape cartridges come in two flavors, single reel and dual reel. Your choice will affect both the capacity and load time characteristics of the system.

The single reel tape cartridge has all the tape on one reel - the take-up reel is inside the tape drive unit. That saves space and cost in the cartridge, providing better alignment of the head assembly. The drive is larger - since the take-up reel is there - but the cartridge has more capacity per cubic inch because there is no empty space in the cartridge.

On the other hand, dual reel has a faster load time, which is the time from when a cartridge is inserted until the tape is ready for read or write of data. Single reel takes about one minute. Dual reel midpoint load is faster and can load in four seconds. In midpoint load, the cartridge has the take-up reel included and is always half empty. Put simply, choose single reel for capacity and dual reel for performance.


A guide to tape formats

Choosing the right combination of tape formats helps calculate a tape library's TCO.
TECHNOLOGY VENDOR EXAMPLE TAPE CAPACITY (NATIVE) SUSTAINED TRANSFER RATE (NATIVE) INTERFACES SUPPORTED FORM FACTOR MEDIA AUXILIARY MEMORY IN CASSETTE
QIC HP 4GB 0.6MB/s IDE 3.5 MP No
4mm DAT Helical HP SureStore 12,20,33GB 3MB/s SCSI 3.5 MP No
4mm Linear Packet Exabyte/Ecrix VXA 12,20,33GB 3MB/s SCSI,1394,IDE 3.5 AME No
8mm DAT Helical Exabyte
MammothL
Mammoth2
14GB
60GB
2MB/s
12MB/s
SCSI,
Fibre
3.5
3.5
AME
AME
No
No
8mm Helical AIT1 AIT2 Sony
SDX3/400
SDX500
25,30GB
36,50GB
3MB/s
6MB/s
SCSI
SCSI
3.5
3.5
AME
AME
64k MIC
64k MIC
DLT Linear Quantum
DLT7000
DLT8000

34GB
40GB

5MB/s
6MB/s

SCSI
SCSI

5.25
5.25

MP

No
No
DLTI Linear Benchmark
VST 80

40GB

3MB/s

SCSI

5.25 half height

MP

No
LTO Linear HP
Ultrium215
Ultrium230
10,30,50GB
10GB
10GB

7.5MB/s
15MB/s
Fibre-IBM
Ultra 2
SCSI

3.5
5.25

MP
MP

32k CM
32k CM
SDLT Quantum 110GB 11MB/s SCSI 5.25 AMP No
3490E Linear STK 9490
Timberline
.8GB
1.6GB
6MB/s ESCON
SCSI
5.25+ MP No
3590 Linear IBM
Magstar
10,20,40 GB 9MB/s or 14MB/s ES/FICON,
SCSI, Fibre
5.25+ MP No
9840 STK
Eagle
20GB 10MB/s
19MB/s
ESCON
SCSI, Fibre
5.25+ MP No
Half-inch Helical STK SD3
Redwood
10,25,50 GB 11MB/s ESCON
SCSI
5.25+ MP No
Source: Evaluator Group

Recording technology. There are many ways that data is actually written onto the tape. The different formats have different characteristics in density, speed and reliability. Linear parallel recording writes multiple tracks simultaneously. Linear serpentine writes data in one channel along the tape and reverse and writes in the other direction on another channel. The cost of writing in serpentine is less expensive than linear parallel due to not having to have an expensive, parallel tape head. A third technology - helical scan recording - is similar to that used in consumer VCRs. Data is actually recorded in stripes at an angle across the tape by the rotating head.

Software support. Tape manufacturers maintain a close relationship with backup/recovery ISVs to ensure compatibility with their new products. A new tape product or library is not ready for use in the market until the ISVs port compatibility with their software and the popular operating systems.

Footprint. Tapes drives follow the computer industry trend demanded for a smaller footprint to reduce the real estate required for the device. Small form factor is key in lower end systems. Tape drives for midrange have been mostly in the 5.25-inch format, while enterprise class drives are larger. The 3.5-inch format is becoming popular in new midrange models to increase the number of drives in a library or to fit in an internal system bay.

Interconnects. There are many ways to connect tape drives to the system platform or host library. SCSI was the popular standard, but native Fibre Channel (FC) connections are increasing in importance, due to the popularity of network-attached storage (NAS) and SANs. Enterprise systems still use ESCON and the newer, faster FICON.

Auxiliary memory. Newer generation tape cartridges have non-contact, non-volatile memory chips that can be read by an RF reader. Metadata of where and what is stored and key diagnostics on these chips can provide vital information prior to loading the tape, increasing the degree of automation libraries can perform. Unfortunately, there is no auxiliary memory standard, which will frustrate automation providers and users until vendors get together and do the right thing. The new generation of tape drives or "super drives" such as LTO, Super DLT, AIT3 and Super AIT will have a dramatic affect on library automation (see "How new generation tapes change libraries" below). The increased cartridge capacity and transfer rate performance of super drives will increase the density and reduce the size of libraries, which will need fewer slots and less media.. That will help maintain tape's 60-fold cost advantage over disk, while providing faster data access to enhance applications


How new generation tapes change libraries

The latest generation of tapes will have a dramatic effect on library automation as they add capacity and capabilities to libraries.
TECHNOLOGY VENDOR EXAMPLE TAPE CAPACITY (NATIVE UNCOMPRESSED) SUSTAINED TRANSFER RATE (NATIVE) INTERFACES SUPPORTED FORM FACTOR MEDIA AUXILIARY MEMORY IN CASSETTE
8mm Helical Sony
AIT-3
100GB 12MB/s IDE
SCSI, Fibre
3.5 AME 64k MIC
LTO HP, IBM
Seagate
Ultrium
200GB 20-40MB/s SCSI,
Fibre
3.5
5.25
MP 32k CM
SDLT 320 Linear Quantum
Super DLT
160GB 16MB/s Ultra2,160
SCSI
5.25 AMP No
S-AIT Helical Sony 500GB 30MB/s Ultra SCSI
160, Fibre
5.25 AME R-MIC
VST (DLT1) Linear Benchmark
VST80
80GB 6MB/s SCSI 3.25 MP No
Source: Open Systems I/O Evaluator Group, Chapter 4, 2002.

On the downside, super drives are not designed for the high duty cycles of the enterprise market. Enterprise class drives like StorageTek's 9840/9940 and IBM's 3480/90 should be used where duty cycles exceed 50%.

Tape libraries are great in SANs
Currently, the killer application for SANs is backup/recovery. Today, most SANs are connected by FC. The FC option on a tape drive costs more than SCSI. FC tape drives are needed in high-end library applications. Until the FC option cost comes down, it's less expensive to use routers and bridges to support multiple SCSI drives in a library. However, it is better to use native support than convert. Also, care must be taken to address the FC topology. For example, special ports or switches are needed that support arbitrated loop.

When using storage resource management software, make sure it's compatible with your tape library and HBAs, switches, and operating systems.Keep in mind that testing should be down to the firmware level. Don't mix and match - use tested, validated configurations. Once you arrive at a stable configuration, your tape library can aid in:

Serverless backup. Most backup software ISVs allow the tape drive to be the initiator in the SAN. Backup is much faster through a host and doesn't degrade the network.

Disaster recovery. SANs can be bridged or linked over large distances by iFCP, FCIP and soon, iSCSI. Even if you have redundant RAID, snapshots to a tape library are required as there are too many single points of failure such as software corruption, viruses and sabotage. Using the right backup software, a SAN can also split or duplex the write to a remote location.

SAN management and virtualization. Used to reduce the complexity and proprietary nature of SAN solutions. Tape, disk and memory strengths are blended and used proportionately to reduce cost. Tape is represented as an image on disk and then tapes are completely filled using volume stacking. StorageTek and IBM are the leaders in this market. SAN virtualization will become important.

Tape striping or redundant array of inexpensive taped (RAIT). This approach works for high bandwidth applications such as telemetry and seismic data. Currently, RAIT applications don't share data well in a SAN, due to EOT and other specific differences. Specialized software will be an important part of your tape library selection for this application.


Popular libraries and tape drive support
Cartridge
slots
ADIC ATL Exabyte HP IBM Overland Data Inc. Storage Technology Inc. Qualstar
20 S100 DLT LTO SDLT M1500/SX DLT LTO SDLT 221L LTO 230D DLT 2/20 DLT LTO 7337 DLT 3570C/XL Non DLT LXM, NEO DLT SDLT LP2/3 AIT L20 DLT DLT1 LTO SDLT 4420 AIT 6220 DLT DLT1 SDLT 8222 LTO
40 S100 DLT LTO SDLT AIT M1500x2/sx DLT LTO SDLT P1000 DLT SDLT 430M Mammoth 4/40 DLT LTO 3583 LTO 3575L06 3590 NEOx2 DLT SDLT L40 DLT DLT1 LTO SDLT 4440 AIT 64xx DLT 6430/60 DLT1 SDLT 84xx LTO
80 S100 DLT LTO SDLT AIT M1500x3/x4 DLT LTO SDLT P1000,7700 DLT SDLT A2000 AIT X80 Mammoth 6/60 DLT LTO 3583 LTO NEOx3 DLT SDLT L80 DLT DLT1 LTO SDLT 4480 AIT 64xx DLT 6460 DLT1 SDLT 68120 SDLT 84xx LTO
180 S1000 DLT LTO SDLT AIT M1500x9 DLT LTO P2000 DLT LTO SDLT P4000 DLT LTO SDLT X200 Mammoth 6/140 DLT LTO 10/180 9840 DLT LTO 3584-227 LTO 3494 3590 3575L18 3590 NEOx7 L180-140 9840 9940 DLT LTO SDLT 412180 AIT 68120 DLT SDLT 88132 LTO
700 S1000-390 /698 LTO S1000-390 SDLT S1000-418/ 668 DLT S1000 AIT S10K AIT DLT LTO SDLT A6000 AIT P7000 DLT LTO SDLT P4000 DLT LTO SDLT   20/700 9840 DLT LTO 3585-227/ 667 DLT LTO 3494 3590   L700-396/ 678 9840 9940 DLT LTO SDLT 412600 AIT 68240 DLT SDLT 88264 LTO
Source: Evaluator Group

This was first published in June 2002

Dig deeper on SAN management

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

-ADS BY GOOGLE

SearchSolidStateStorage

SearchVirtualStorage

SearchCloudStorage

SearchDisasterRecovery

SearchDataBackup

Close