Managing and protecting all enterprise data


Tape's new love affair with disk

The marriage of tape and disk has spawned a new class of virtual tape products that promise faster, cheaper backup and recovery.

Alternative disk and tape systems
Not all users can afford a high end, tightly integrated disk and tape system. If budget dollars are scarce, consider these alternatives:

Copan Systems Revolution 200T: The Revolution 200T offers up to 896 drives and 224TB of capacity in a single cabinet. Using its Massive Array of Independent Disks (MAID) technology, it powers on and spins up disk when needed, limiting wear and tear on SATA drives. The Revolution 200T emulates multiple tape libraries and tape drives (LTO, DLT, SDLT and 9840), and can present 56 virtual tape libraries or drives when fully configured.
FalconStor's Virtual Tape Library: This software product runs on an off-the-shelf Intel-based Linux server, connects to any SAN or direct-attached Fibre Channel, SCSI or SATA disks and presents the disk as virtual tape drives/libraries. It can emulate a variety of tape drives (SAIT, SDLT or LTO) and libraries (ADIC, HP, IBM, Sony, StorageTek), and works with most backup software packages. It can be used in conjunction with existing tape libraries to export the virtual tape drives/libraries to physical tape cartridges.
Overland Storage REO 4000: This box offers up to 8TB native capacity in a 2U form factor with Gigabit Ethernet iSCSI and optional Fibre Channel connectivity. The virtualization software on this device allows the unit to present its disk as either disk or tape images or a combination of both. The disk may be configured as JBOD, RAID 0 or RAID 5, and the unit supports up to a total of 16 LUNs or virtual tape drives. Pricing starts at around $12,500 for a 2TB system. Overland also offers a smaller unit, the REO 1000, with a MSRP of $5,000.
Spectra Logic's Spectra RXT: This is a portable RAID Serial ATA disk pack. Depending on how it's configured, it can operate either as a standalone solution or it can be inserted and managed in a standard Spectra T950 tape library drive and media bay. Each RXT can hold up to 400GB of SATA disk and offers RAID levels 0, 1 and 5. The RXT emulates LTO-2 tape drives and can be used with most backup software packages.
If you can't beat 'em, join 'em. That's the rallying call embraced by tape library vendors as they respond to the steady encroachment of disk on their core tape business. Front-end software transforms disk into virtual tape drives and policy-driven backup software handles the data migrations from the disk to the tape and back again. This new climate of coexistence between disk and tape gives users a powerful new tool for improved data protection, high availability and information lifecycle management (ILM). There are two compelling reasons to deploy disk in front of tape libraries: better backup and recovery times and lower costs.

Jumping on the cheaper disk bandwagon, tape library vendors now provide their own ATA disk arrays with front-end I/O controllers running tape virtualization engines that present themselves as virtual tape drives to backup software. The new virtual disk interface gives storage administrators the ability to transparently choose between tape and disk as the primary backup target and to migrate data between the two using policies nestled within their backup software. The integration of disk and tape also frees users to replicate data during off-peak times to off-site disk arrays for disaster recovery purposes, and then transfer the data to tape to satisfy vaulting requirements.

Despite the appealing benefits that closely integrated disk and tape offer, users first need to verify that the vendor's disk array tape virtualization product is compatible with their backup software and storage environment. For example, Spectra Logic Corp.'s disk-based RXT appliance only emulates LTO-2 tape drives, while IBM Corp.'s B20 Virtual Tape Server (VTS) systems only provide a 3490 tape drive target. The limited tape selection could be a compatibility issue in shops where either SDLT or SAIT are the main tape media used. For vendors targeting Unix and Windows environments, the trend increasingly is to support LTO as the standard virtual interface.

IBM/Tivoli targets its disk/tape product for storage environments with mainframe, Unix and Windows operating systems. IBM offers its disk-based B series line of VTS, along with its 3590 tape systems. The VTS and 3590 tape systems can be monitored and configured using Tivoli's Storage Resource Manager (SRM) and Storage Manager (SM) tools, which enable policy-driven data management to move data between the virtual and real tape subsystems.

Storage Technology Corp. (StorageTek) also offers a range of tape libraries and disk subsystems for MVS (mainframe) storage environments. StorageTek's L, SL, PowderHorn and TimberWolf tape libraries work in conjunction with its B, D and V series of disks under the umbrella of the company's Virtual Storage Manager (VSM) software. The VSM software emulates T9840 and T9940 tape drives. In addition to the VSM software, StorageTek also offers Storability Software Inc.'s Global Storage Manager (GSM) that reports on both the tape libraries and disk arrays.

Recently, a number of storage vendors have partnered to fill gaps in their disk/tape offerings. Advanced Digital Information Corp. (ADIC), for example, filled two holes in its product lineup through its recent alliance with EMC Corp. ADIC offers neither disk arrays that compete in the high-end and midrange areas of the market nor storage resource management software that works outside of their product line. EMC fills these gaps with its high-end and midrange arrays and its Control Center SRM software. EMC also provides an enterprise tape library product to round out its ILM strategy.

Inside tape virtualization software
Tape virtualization software offers users some distinctive benefits over existing tape library solutions.

Tape virtualization software allows the front-end I/O controllers to appear as tape drives to backup software. Software is available to emulate almost any tape drive on the market, though the selection varies by vendor. ADIC's Pathlight VX presents LTO tape drive targets; IBM's Virtual Tape Server (VTS) offers only 3490 tape drives; StorageTek's VSM provides T9840 and T9940 tape drives. The most versatile product on the market appears to be Copan Systems Revolution 200T (see "Will MAID clean up messy tape backups?" in the July 2004 issue of Storage), which emulates a number of popular tape libraries and drive types.
Existing tape drives can present only one image to backup software, so backup jobs must be queued up. Virtual tape interfaces alleviate this bottleneck by allowing users to create and present a new tape drive interface to each backup job. High-end systems, such as those from IBM and StorageTek, are able to support up to 256 virtual drives.
Backup performance should increase substantially. Some virtual tape products reach speeds up to 500MB/sec because backing up to disk eliminates the need for tape mounts or backhitches.
Tape virtualization software creates volumes out of the disk pool that appear as tape cartridges to backup software. This allows backup software to manage and expire virtual tape volumes in the same manner it handles real tape cartridges.

The influx of ATA disk
While companies such as IBM and StorageTek have offered integrated tape and disk products for the mainframe for years, it's only with the introduction of low-cost ATA disks that this solution becomes affordable to the masses. The quality of disk needed for backups is generally lower than what's required for day-to-day operations, and the price per gigabyte is significantly lower than Fibre Channel (FC) disks.

Despite this allure, users must still verify some basic facts about an ATA array:

  • Does it offer tape virtualization software?
  • If it offers tape virtualization, what types of tape drive interfaces does it present?
  • Will it support the I/O for your environment?
  • Does the array support RAID? If so, in what configurations?
  • What software is used to integrate the array with tape libraries?

The new generation of ATA disk arrays uses either virtualization software provided by the tape library vendor or third-party software. Storage administrators should pay particular attention to front-end virtualization products. While every vendor's disk array offers disk virtualization software, not every product provides front-end tape virtualization software.

For instance, EMC's Clariion CX array is available in an assortment of configurations that can be packaged with different front-end disk virtualization software. If purchased from EMC, the DL700 ATA disk arrays will ship with FalconStor Software's IPStor virtualization software. However, if purchased from ADIC, the same back-end EMC disk array will ship with ADIC's virtualization software, obtained as part of its acquisition of V-Stor in 2002, and ADIC's I/O controllers.

Other virtualization products from vendors such as Candera Inc. or IBM don't present virtual tape drives or cartridges, but rather virtual disks or LUNs from pools of storage; these don't have functionality to allow backup software to see virtual tape drives or cartridges.

While users can obviously implement backup software or normal operating system utilities to store data on this type of disk, they shouldn't expect the same level of functionality that they'll get from arrays that have front-end tape virtualization software.

For example, ADIC, IBM and StorageTek use software that migrates data transparently between the disk array and the tape library. This provides a number of important benefits. First, the data transfer from the virtual tape disk to a physical tape can occur at any time. Secondly, the identifier assigned by the backup software to the virtual tape moves with it to the physical tape, so when the tape is ejected, the backup software isn't aware there was a virtual tape involved at any point in the process.

Users also need to weigh variables such as reliability, support and warranties on arrays with ATA disks. Remember that disk drives, power supplies and internal components fail, or worse, have intermittent problems. ATA disk is cheap for a reason: Qualstar Corp. finds that ATA disks experience failures at rates six times greater than SCSI disks and have a mean time between failure (MTBF) of 600,000 hours. While this exceeds the 400,000 hour rating of LTO, SAIT and SDLT tape drives, the loss of a tape drive won't result in the loss of data, while the loss of a disk drive could. In fact, Dell Inc. finds that a standard RAID-5 configuration in a Dell Power Vault 660F/224F array with 14 SCSI drives carries a 38% chance of data loss over a three-year period. If you factor in the greater likelihood of ATA drives failing vs. SCSI drives, the probability of data loss using ATA drives in some RAID-5 configurations increases substantially.

Virtual tape subsystems

Tying disk and tape together
To get their disk/tape arrays to work, in most cases, vendors just rework their own storage resource management software or set up some customized policies in the backup software users already own. However, there are exceptions. Tricia Jiang, an IBM Tivoli technical representative, says, "TSM views disk, tape and virtualized disk as a storage pool on which it can store data based upon policies set by the administrator. It's agnostic in its view of the storage it manages, so if a user has EMC Symmetrix and IBM FAStT disk systems with a Spectra Logic tape library and an ADIC Pathlight VX virtual tape subsystem, it does not care."

The integration of disk into tape libraries combines the best qualities of both disk and tape in a single backup system. These new disk/tape libraries offer users the ability to meet immediate backup and recovery needs and address their longer term compliance requirements, while setting the stage for off-site data replication and vaulting. Users should begin to identify methods to incorporate this integration of tape libraries and disk into their environments. Failure to do so will ultimately mean organizations will pay the price in terms of decreased efficiencies and higher costs both to maintain the availability of their data as well as to recover it.

Article 15 of 20

Dig Deeper on Data storage strategy

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.

Get More Storage

Access to all of our back issues View All