Tape: Alive and full of options

Sure, disk is going down in price, but don't underestimate the value of tape technology. We look at what's available in tape libraries for large-scale backup.

This article can also be found in the Premium Editorial Download: Storage magazine: What are the real benefits of data storage management software?:

Despite major advances in disk technologies, tape continues to be the mainstay of most companies' backup efforts, despite being viewed to some users as throwback technology. New capabilities such as higher tape densities, the ability to manage different tape mediums in one library and increasingly higher rates of reliability are keeping tape the most cost-effective storage medium.

Sony introduces AIT drive for PCs
Sony Electronics recently released an external Advanced Intelligent Tape (AIT) drive for PCs and laptops with USB and i.LINK/FireWire (IEEE 1394) connectors for most PCs and Macintosh computers, as well as an internal ATAPI version for use in entry-level servers and workstations. With maximum compressed capacities from 91GB to 130GB and maximum native data transfer rates from 4MB/s to 6MB/s, Sony's new StorStation AIT drives start at $800. Sony's new external AITe90-UL drive will include a complimentary copy of IVision Software Inc's ISafe backup software for Windows.

Rick Luttrall, director of marketing for Hewlett-Packard's Nearline removable storage line of products, uses a rule of thumb to help customers determine the tape resources they need. For every 1TB of disk storage, a company should have three to five times as much tape available. While not a hard and fast rule, Luttrall finds that a 3:1 ratio gives an organization enough tape for about a month's worth of unattended backup, while the 5:1 ratio will give about a quarter's worth of unattended backup.

Assuming Luttrall's numbers hold true (see "How much tape do you need?"), these ratios should serve as a wake-up call to a silently growing cost in organizations: tape media. While an organization may upgrade their disk subsystems to take advantage of higher capacities and lower costs per megabyte, it usually fails to give the same attention to their tape technology, relying on what's worked in the past. This failure to upgrade can mean escalating costs.

For instance, older DLT 4000 tape cartridges hold 20GB of data natively and 40GB compressed, back up at a rate of about 10GB/hour natively, 20GB/hour compressed and cost $70 to $100 per cartridge. Contrast this with the more current SDLT 320 tape cartridges which hold 160GB of data natively or 320GB compressed, back up at 57.6GB native, 115GB compressed and cost about the same ($100). A move to a more current tape media can save money and shorten the backup window (see "Tape media formats").

Tape library clusters
Today's tape libraries can scale to manage terabytes of data and hundreds of tapes in one frame. In fact, an increasing number of tape library vendors are clustering their individual frames to create one large tape library unit that can scale to manage petabytes of data and thousands of tapes.

The Advanced Digital Information Corp. (ADIC) AML/2 illustrates this clustering functionality. The AML/2 comes in QuadroTower modules, nine of which may be configured as a single unit which scales to hold more than 5PB of data and 400 tape drives. It also hosts and manages multiple tape mediums simultaneously within a single frame. The 20 mediums it supports range from DLT, SDLT and LTO tape formats to 3590 and 9840 tape mediums found in high-end mainframe data processing environments.

Ever try to restore data
from 20-year-old tapes?
Tape's longevity--one of its greatest strengths--is also its greatest drawback. Managing and tracking tapes that go back decades, much less restoring the data on those tapes can become a logistical nightmare. To circumvent some of this pain, companies should follow one of two policies: First, as data is needed from older tapes, it's recalled and migrated to newer media. And secondly, even if the data is never recalled, as time passes, the data is migrated to a newer generation of tape to ensure the data remains recoverable for the foreseeable future.

StorageTek PowderHorn 9310 tape library is also a highly scalable unit. A single 9310 library may be configured to hold 1.2PB of data supporting 80 of StorageTek T9940B tape drives. While these 9310 libraries lack the wide range of multimedia support ADIC's AML/2 offers, you may configure 24 tape libraries into a single cluster that may hold over 57PB of data and 960 tape drives.

For organizations looking to consolidate their diverse tape mediums into one tape library and still use the tape medium, a product like ADIC AML/2 makes a lot of sense. In addition, most current tape drive technologies remain compatible going back a few generations. For example, the most current SDLT tape drive in the AML/2 reads information from DLT tape cartridges that go back two or three tape generations. StorageTek's tape drives support any combination of their past five generations of tape media technologies down to the individual slot all in the same library.

While these two models have some impressive features and scale to support the highest data processing environments, many other models exist that fit into enterprises of various sizes such as the IBM 3584 LTO UltraScalable Tape Library. It scales to over 2,400 tape cartridges, allowing for up to 72 tape drives in a single unit. Each of its tape drives--using the latest IBM LTO Ultrium 2 technology--support native streaming data rates of 35MB/s. The unit itself can scale up to nearly 1PB of data in a single frame assuming the use of 2:1 compression.

How much tape do you need?
DLTtape.com gives some practical formulas to assist companies in determining how much tape they need depending, of course, on the tape capacity your company has standardized on.
TAPES FOR STORAGE
Xs = D * T * S * R + N
Xs = number of tapes per year for storage
D = number of backup drives
T = number of tapes in the media set
S = number of sets in rotation
R = number of rotation schedules per year
TAPES FOR ARCHIVE
Xa = T * S * A
Xa = number of tapes per year for archiving
T = number of tapes required to copy each server
S = number of servers
A = number of archive sets per year
TAPES FOR DISASTER RECOVERY
Xr = T * S * R
Xr = number of tapes per year for recovery
T = number of tapes per server (from above)
S = number of servers (from above)
R = number of disaster recovery sets required
TOTAL ANNUAL TAPE REQUIREMENTS
X = Xs + Xa + Xr + R
X = total number of tapes per year
Xs = number of tapes per year for storage (from above)
Xa = number of tapes per year for archiving (from above)
Xr = number of tapes per year for recovery (from above)
R = number of new replacements for tapes retired

Beware the backhitch
IBM's senior program manager, Bruce Master, notes that the streaming data rates of these new tape drive technologies have gotten so fast that the backups can run faster than the rate at which they receive data. This creates a new problem during the backup period called backhitch.

A backhitch occurs when the flow of data onto a tape during a backup gets interrupted. When this occurs, the tape drive can't immediately stop. Instead, it must slow down and stop the tape, back the tape up and reposition the tape at the point where the data flow was interrupted. The time to perform a repositioning is short, but frequent backhitching can significantly degrade performance.

To counter backhitching, IBM incorporated a new technology called Digital Speed Matching into its LTO Ultrium 2 tape drives, which lets the drive adjust as closely as possible to the speed and flow of the incoming data. This decreases backhitches and improves throughput performance, says IBM.

Other vendors have encountered similar issues and implemented their own solutions. StorageTek provides an adequate buffer size to mask the backhitch. On their high-end 9840B and 9940B tape drives, they incorporate 32MB and 64MB buffers to mask this backhitch. DLT and SDLT tape drives use a methodology similar to LTO's methodology called adaptive cache buffering which automatically adjusts the DLT or SDLT tape drive's transfer rate to match the host system's data rate.

Reliability, density and scalability
Hewlett-Packard's enterprise class ESL9595 tape library typifies the three requirements of nearly every tape library: reliability, density and scalability. For reliability, the ESL9595 tape drives connect directly into the backpane which is part of the intelligence of the unit. While other tape library manufacturers offer hot-swappable tape drives, because ESL9595's tape drives plug directly into a backpane, no cabling must be disconnected to remove the drive, which reduces the risk associated with exchanging failed tape drives on the fly by isolating the other drive in the library from this change.

How often tape drives fail depends on the tape drive. The ESL9595 supports both LTO and SDLT tape drives. The LTO tape drive has the longest head life of any tape drive on the market--coming in at 60,000 hours--or nearly seven years. This contrasts with the SDLT tape drive which rates at 30,000 hours or about three and a half years. For those looking beyond the LTO and SDLT mediums, Qualstar rates the life of the heads of their AIT tape drives at over 50,000 hours, or nearly six years.

Density is also important as a tape library function. HP helps customers maximize capacity per square meter on their data center floors, without taking up precious floor space. To achieve this, a fully configured ESL9595 may hold 595 SDLT tape drives--or up to 190TB compressed--in over just one square meter of floor space.

Spectra Logic's Spectra 64K holds nearly 168TB of compressed data using 645 AIT tape cartridges in under one square meter of floor space. On the LTO side, StorageTek's single frame L700e scales to 271TB compressed using 678 LTO Gen 2 tape cartridges in a single library on about 1.75 square meters of floor space.

Scalability is the final must for most tape library products today. The pass-through mechanism (PTM) in the HP ESL9595 reflects this third necessity. The ESL9595's configuration allows up to five ESL9595 tape libraries to be joined together into one multiunit tape library system. The PTM technology helps give this library the additional functionality to allow tape cartridges to be shared between any of the libraries connected in its multiunit configuration.

To what degree tape libraries scale vary according to vendor. StorageTek L700e enables two of these units to be configured as one logical unit. Quantum's ATL P7000 equals HP by allowing five of their libraries to function as one logically scaled unit. ADIC ATL/2 enables nine of their modules to be configured as one, while StorageTek 9310 PowderHorn grows to manage 24 units as one logical unit.

Tape Media Formats
DLT: Ranges in capacity from 10GB native/20GB compressed (DLT III) to 40GB native/80GB compressed (DLT IV). The older DLT III tapes have a shelf life of more than 20 years, while the latest DLT IV technology extends the life of the tape to more than 30 years. All DLT tapes are readable by the latest SDLT tape drive technology.
SDLT: Represents the current generation of DLT tapes. The SDLT 220 has a capacity of 110GB native/220GB compressed while the latest SDLT tape cartridge--the SDLT 320--holds 160GB native/320 compressed. The SDLT 320 media supports compressed transfer rates of up to 32MB/s and SDLT tape drives are capable of reading, but not writing to, older DLT tape media. SDLT tape media also have a life expectancy of more than 30 years. DLT and SDLT have combined to sell over 90 million cartridges and two million tape drives.
LTO Ultrium: Jointly designed by HP, IBM, and Seagate and is a relative newcomer to the tape space, first being introduced in 2000. The most current product shipping supports 200GB native/400GB compressed, supporting a tape drive throughput speed of 35MB/s native and 70MB/s compressed. IBM estimates a shelf life of up to 30 years for LTO Ultrium tapes and has already shipped over five million LTO cartridges in just a little over three years.
AIT: First introduced into the market in 1996 with three formats currently available on the market: AIT-1, AIT-2 and AIT-3. AIT-3 natively supports drive capacities of 100GB and with a higher than average industry compression rate of 2.6:1, gives it a compressed capacity of 260GB. It offers a transfer rate of 12MB/s native, 31MB/s compressed.
Next-generation tapes: Brings still higher capacities and faster throughput to the marketplace. The SDLT 600 tape cartridges are due out Q4 of 2003 with compressed capacities of 600GB and 64MB/s throughput rates. The 2006 forecast for the SDLT 2400 anticipates over 200MB/s throughput and 2TB of data compressed on a single tape media. Both LTO Ultrium and AIT include similar high transfer rates and media storage forecasts in their roadmap over the next one to three years. For instance, SAIT-1--released February 2003 by Sony--became the first tape media to break the 1TB barrier with the ability to store 1.3TB compressed (assuming a 2.6:1 ratio) on a single tape. Sony anticipates tape library solutions from vendors such as Qualstar will possibly be available as soon as spring 2003 to support this new format.

Where's tape headed?
Now that I've listed many of the technical advances in today's tape libraries we should address that age-old argument: Tape is dead. Nothing could be further from the truth.

IBM's Bruce Master believes that the cost of tape still beats the cost of disk by a long shot. Despite the recent price drops of low- cost disk to under a penny a MB, tape media is still less expensive by a wide margin. Future IBM tape technology, he says, could hold a terabyte of data on just one tape, which equates to about ten cents a gigabyte. HP's Luttrall adds that no matter how much the cost of disk drops, disk cost will still exceed the cost of tape by ratios ranging from 20:1 to 50:1.

Yet none of the tape vendors dispute the value of disk in the next generation of backup solutions. In fact, HP sees disk as a critical piece of their multilevel data protection strategy and views technologies such as ATA arrays as being complementary solutions to tape, serving as a disk cache in front of tape to stage the data prior to being backed off to tape. Some of the tape libraries on the market have already capitalized on this idea. In 1997, IBM introduced their Virtual Tape Server that uses disk to present a virtual tape subsystem to the host system. While primarily in use on the mainframe, it's has crept into a few open systems environments and IBM foresees this technology growing in the future.

Another area where tape outshines disk is in offline storage, because tape is portable and removable. HP's Luttrall says that when data remains online as it does in the case of disk, it remains vulnerable to corruption. Tape, on the other hand, can be moved and vaulted. While this introduces some management costs and a potential for tape destruction or loss, it also protects the data from human error--accidental or malicious. It also allows the data to be moved offsite in case of outage or disaster where the primary data facility resides.

Finally, tape just flat out outperforms disk in its ability to read and write data sequentially. Many of today's tape libraries may host anywhere from 16 to 400 tape drives. Couple this number of tape drives with new tape drive technologies now supporting speeds of over 30MB/s and it adds up to backup speeds that disk simply can't match.

Consolidate or eliminate?
Should you consolidate all your tape libraries down to one or standardize on a tape library that only handles one tape format? A case can be made for each approach.

The argument for putting all of the tape mediums into one tape library sounds compelling: Place all your tape mediums into the one library and then verify compatibility with backup software products. This contrasts with standardizing on one tape medium. However, any library that reads multiple tape formats has its downsides: It's more expensive and has more management complexity. If you throw AIT or SDLT tape drives into the mix, your backup speeds and times will vary and you'll always be asking which application is backed up on what tape format. In addition, each of the tape mediums will have different life expectancies and handling requirements, whereas with only one tape format, you'll only have--at most--two or three variables.

In either scenario, you have a major management headache. In the case of a single tape format, there's the pain and expense of moving data to that solution. In the case of multiple tape mediums, you may end up chasing your tail trying to figure out which application is backed up by what type of tape media. The bottom line: The costs of the actual physical devices probably evens out and the ability to manage the environment becomes the key long-term differentiator.

This was first published in April 2003

Dig deeper on Storage vendors

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