Tape libraries central to intelligent tape: tape proves its value in data center, disaster recovery.
Tape is no exception. Tape storage vendors are constantly increasing tape speed and capacity, and are also placing increasing amounts of intelligence and automation into Fibre Channel tape libraries. But is this intelligence being used as well as it could be?
Tape benefits and suffers from the idea that it only does backup and recovery. This is its main job--tape's claim to fame is cost-effectiveness, portability and reliable media. But tape has recently suffered hits from disk backup competitors as users, vendors and the press ask either/or questions: Should everyone back up to disk? (No.) Should some applications back up to disk? (Yes.) Is tape going away? (No.) Is disk going to replace tape? (Not for a long time.)
Why? Since companies can back up to disk (and library developers themselves are putting in disk front-ends), why go with tape? The reason is portability, which is still vital in disaster recovery schemes. Senior analyst Steve Duplessie of the Enterprise Storage Group, commented, "Using inexpensive disks to recover from nondisasters makes sense. But companies will still put a tape backup in a bunker in Provo, Utah, just in case a bomb goes off."
But portability isn't the single reason administrators are still using tapes. IT does get tired of shifting tape around, and many of them long for the ease of disk-based backups. However, the reality is that archival storage still revolves around tape; it retains a huge installed base, remains much less expensive than disk (even serial ATA) and its portability provides high levels of protection and security. (It's really hard for hackers and viruses to attack off-site tape.) These qualities alone will keep tape firmly planted in data centers, and ongoing advances in tape, especially automated management tasks and disk/tape hybrids, will keep tape an important element of the storage networking equation.
Tape libraries are central to intelligent tape. Just as centralized disk arrays are more cost-effective than direct-attached disk, tape libraries are more cost-effective than individual tape drives on individual servers. Since libraries use centralized media, tape owners reduce tape handling and experience more efficient operations, because they don't have to manually switch out tape as often as in smaller libraries and autoloaders. Because libraries are critical for efficient backup operations, developers are placing intelligence in the libraries as well as in individual drives. Intelligent tape includes features like:
Intelligent processor: Most tape library vendors concentrate intelligence in the processor, which should have high speed and vast I/O capability. Recent advances like this allow highend library developers to dispense with external library controllers.
Front-end disk cache: Many enterprise library makers have combined tape and disk subsystems by placing a disk array on the front-end of the tape library. The disk is configured to look like tape so backup applications will write to it, but it captures backup data quickly and archives it at leisure.
High-end tape library: Libraries house many disk drives for high capacity, boasting highly reliable tape drives, robust robotics, and SCSI and FC connectors. Ideally, it should be neutral on the drive and media types it can hold. Several library developers have included internal partitioning capabilities.
Management software: Includes internal management software for system management, external sharing and integration, and diagnostics for SAN management applications such as SRM.
Standards-based: New tape libraries should be SMI-S compliant. SNIA's SMI-S standard is key to for managing communications and integration within heterogeneous storage networks.
Several advances have kept tape an important part of the data center, including more sophisticated management software, disk/tape hybrids, tiered storage architectures and virtual tape.
Emerging library management software improves diagnostics, alerts, threshold requirements and drive/media management. Advanced Digital Information Corporation (ADIC) and Quantum, for example, have recently introduced native management software tools on their tape libraries and drives. ADIC's Scalar i2000 can alert administrators to potential problems, partition itself into logical libraries and carry out sophisticated self-diagnostics. Quantum's DLTSage, which it introduced last summer and is still pushing hard, runs proactive diagnostics like critical threshold alerts and error condition warnings.
One of the more interesting developments is a tape library with a disk front-end. In this model, data writes quickly into the front-end disk cache and later writes to the back-end tape pool without impacting applications or backup performance. Policies control the migration to tape: data that is being migrated to long-term archives might pass immediately through the disk cache, while backed-up active data might stay in cache for some time, allowing fast recovery. (Common time periods are a week for daily file backups to 6 months for database data.) The library management software may be the source of the migration policies, or it may work with third-party HSM tools (Hierarchical Software Management).
Tiered Storage Architecture
Even when the tape library lacks a disk front-end, it may work with backup disk drives in tiered storage schemes. Tiered architectures, which are focused around archiving and data protection, start with high performance/capacity disk arrays at the top level. These arrays are the backup target for the application servers. Based on storage policies, migration tools track file characteristics and migrate accordingly to secondary disk arrays, usually made up of ATA disks. (ATA lacks the high performance, speed and reliability of SCSI disk drives, but are cost-effective and work well in secondary archiving levels.) Finally, data migrates from the secondary disk to the tape library, and from there is transported offsite to long-term storage. This setup is also called classes of storage, where firms assign storage targets and infrastructure according to the priority of the stored data.
Using disk for backup applications is not an easy call. Most backup applications expect to find tape at the end of the backup road, and administrators must make changes to drivers and operating system codes if they want their backup to write to disk instead. Virtual tape helps to solve this problem--the technology runs on a diskbased storage target, and makes the disk appear to be tape to the backup application. This allows administrators to avoid changing drivers while easily keeping stored data online. However, it's not perfect; backup applications write sequentially to virtual tape just as if it were real tape. The only way to get around this is take the plunge and make the changes necessary for the backup application to write to disk. This allows administrators to take advantage of disk's speed by using its random access technology.
Tape remains extremely useful in the data center, which has killed the tape-is-dead argument. But tape developers know very well that they are competing against disk-based storage as well as each other. They're directly competing by adding new management capabilities, speed, capacity and ultra fast robotics. They've also adopted the "if you can't beat 'em, join 'em" philosophy by adding disk cache frontends and serving as effective storage layers in tiered environments.
Heh, whatever works ...
RELATED ARTICLE: Intelligent Tape
Arguably, many tape libraries are already intelligent subsystems. High capacity, sophisticated read and write capabilities, intelligent processors and front-end disk caches have made enterprise-class tape libraries a best seller in SANs and data centers. What are some of the high-end library developers doing?
Quantum ATL: DLTSage performs error analysis and predictive diagnostics on its SDLT drives. This helps backups succeed and alerts administrators to upcoming problems with tapes and drives. Quantum also makes the DX30, a tape subsystem with front-end disk. The disk is a virtual tape system and performs as the backup target archiving to the back-end tape system.
ADIC: Scalar i2000 functions include native partitioning into multiple logical libraries, mixed tape drive technology, native Fibre Channel connectivity, advanced performance monitoring and alerts via standard e-mail and pager networks. Pathlight combines virtual tape on ATA disk and back-end tape systems.
StorageTek: Routinely places intelligent services into its tape libraries, which range from workgroup to country-sized silos. They also make the V-series of virtual disk and BladeStore, an ATA disk subsystem. StorageTek will wed tape and disk by putting BladeStore together with its Powderhorn libraries.
IBM: Given IBM's advanced software divisions, it's not surprising that tape management intelligence is highly developed. IBM's 3494 enterprise tape library supports Virtual Tape Server (VTS), which uses virtual drives of RAID-5 disk arrays to cache data and archive it to tape.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Storage Management|
|Author:||Chudnow, Christine Taylor|
|Publication:||Computer Technology Review|
|Date:||Nov 1, 2003|
|Previous Article:||Tiered storage: an idea whose time has come; it's just one driver of the new storage dynamism.|
|Next Article:||Certance announces CP 3100, simplifying disk-to-disk-to-tape for SMBs.|