Tape Feasts On All-You-Can-Eat Data Banquet. And Enjoys.

The magnetic tape industry is building the foundation for the third era in its near 50-year history. The first era began in 1952 when IBM delivered the computer industry's first tape drive and lasted for 35 years. This phase was defined as the era of manual tape handling (mounts & dismounts) and the primary application for tape throughout this period was data backup.

The second era began in 1987 and was re-defined with the successful arrival of robotic libraries providing automated tape services for what had been a very labor-intensive industry. This era was referred to as the Nearline era and was successfully pioneered by StorageTek. Nearline soon became the defacto standard name for the layer of storage between online and offline storage. The resulting benefits from these automated storage services eliminated many of the troublesome growing operational aspects that manual tape faced since its inception.

The third era of intelligent tape storage is beginning. With the world producing between one and two exabytes (lx[10.sup.18]) of digital data every year, the need for mass storage capacity is obvious. Disk and tape are the two most widespread storage devices in use today. It is estimated that nearly 90% of the world's digital data presently resides on removable storage devices, primarily tape, with magnetic disks containing the remainder. Where does tape optimally fit and what architectural enhancements should the tape industry make in order to remain the primary mass storage technology? The price per megabyte purchased for automated tape storage today ranges around 1/25th the price per megabyte purchased of disk storage. In contrast however, tape storage has some limitations compared to disk. Tape access takes seconds versus milliseconds on disk to access the first byte of data and tape can only store and retrieve data sequentially. Since the majority of digital data is accessed relatively infrequently, the need for an improved mass storage system that meets the growing needs of the future has arrived.

Several data storage technologies are positioning to capture significant share of the exploding mass storage and data archive market. Today, tape is the primary choice. To respond to the challenge, the tape industry is looking at several new features and capabilities that improve its value proposition.

1. Price per megabyte. Current recording improvement rates for tape now shows a path to get to 6 gigabits per square inch. Areal densities may ultimately approach 20,000 TPI (tracks per inch) with prices less than $.0005 (five one-hundredths of a cent) per megabyte purchased by the year 2007, staying well below the price per megabyte purchased curves of disk drives. Media prices for new high-capacity tape range from $75 to over $150 per cartridge or about 1/15th of a cent per megabyte for a 100-gigabyte cartridge. Reliability estimates for emerging high capacity tape drives typically range from 200,000 to 300,000 hours between failure.

2. Tape arrays. Technologies borrowed from disk products are increasingly finding their way into tape products such as tape arrays, which can boost performance and capacity while improving data integrity through fault tolerance. Tape arrays are just beginning to arrive and borrow the basic concept from RAID disk arrays. Early tape arrays have been implemented in software but advanced, larger scale implementations may be built in the tape control unit. Tape arrays offer a "multiplying" capability for data rate and capacity while offering striping, mirroring, or striping with parity across tape drives.

An emerging architecture for tape subsystems and certain data transfer-intensive applications, tape arrays allow enhanced data reconstruction and enhanced integrity. Tape arrays may consist of data records or blocks that are striped across (N) cartridges on (N) drives providing data transfer concurrently streaming across multiple tape drives. Data is striped across cartridges and drives in different libraries to provide parallel mounting to occur though mount time is not a major issue in data transfer-intensive applications. Though desirable, an automated tape library is not required for tape array implementation. Mirroring allows two identical tapes to be created in parallel at geographically different locations. As was the case for disk arrays, tape array levels are now being defined and include levels 0, 1, 1/0, 3, and 5. The terms RAIL (redundant arrays of independent libraries) and RAIT (redundant arrays of independent tapes) now commonly describe this concept. Tape arrays address certain throughput int ensive applications and are now appearing for large files and should play an increasing role with streaming video.

3. Capacity/performance. Increasing tape capacity without providing corresponding improvements in tape data rate pushes magnetic tape into more of an archival technology niche. This limits tape's role as a technology that can provide timely access to enormous volumes of data. Therefore, tape data rates and overall throughput capability must keep up with the growth of tape capacity. Further advances in thin media and recording heads are appearing in vendor roadmaps that lead to native (non-compressed) tape cartridge capacities reaching 800GB and data rates up to 100MB/sec over the next five years for both linear and helical scan formats. The ability to store data on tape is important; the ability to retrieve it in a meaningful timeframe has become business critical. "Business resumption is now the killer app for tape".

4. WORM. The use of a logical WORM (Write-Once Read Many), also called write-blocking using encrypted keys in tape subsystems will expand tape usage into applications that were originally targeted for WORM optical disks. With the WORM optical disk having faded as a viable data center storage solution, the tape industry has the opportunity to capture the non-alterable storage market for medical imaging and many government applications.

5. Intelligent cartridges. The intelligent cartridge containing memory chips or "electronic indices" inside the cartridge provides externally sensed information about the cartridge contents enabling more rapid sear h and record locating. Intelligent tape drives will also begin to use infrared external diagnostics to redefine traditional methods of problem determination. The use of midpoint load shows promise for improving initial access times on dual-axis tape cartridges by theoretically reducing the distance needed to find the first byte of data.

6. Virtual. Virtual tape is commonly used today for enterprise (mainframe) systems and will begin to find its way into the Unix, NT, and Linux storage markets. Virtual tape systems deliver better performance, improved utilization of tape cartridge resources, and serve to separate the physical tape technology from the users. Virtual tape subsystems use a RAID disk subsystem that is defined to the enterprise or OS/390 processor as a range of tape transports (not disk devices) also called virtual tape drives. The disk array serves as a front-end cache or buffer for the back-end tape (library) storage.

A tape data set initially written to this array is called a virtual volume. For enterprise applications, the operating system and the tape management system both see this as activity to a real tape volume on a real tape drive. Later, and transparent to the user, the virtual volume is migrated directly from the disk array to the physical tape cartridge on a real tape drive and stored in an automated tape library. By stacking multiple virtual volumes onto one physical cartridge, virtual tape systems utilize cartridge capacity more effectively and have the potential to reduce the number of cartridges in any given application. Reading and writing data from disk minimizes the number of mounts, loading, rewinding, and unloading of cartridges while improving performance. The first virtual tape offerings have been successful in the enterprise market. Now, the vast network storage and midrange markets are being targeted for virtual tape subsystems. Here and unlike the enterprise market, the front-end disk array will need to appear to the server operating system as a disk device, not a tape drive. Data recovery from a virtual tape storage system (a disk array) using advanced techniques such as snapshot, incremental, differential, or forensics can significantly shrink data recovery times.

7. Network storage. Storage virtualization offers several advantages for attaching intelligent tape libraries to SANs, masking the numerous tape formats from different servers while providing heterogeneous access for mass storage. In addition, tape automation provides higher performance backup, recovery solutions for NAS subsystems, which often don't effectively address the backup and recovery issue. While tape itself addresses only sequentially accessed data, efforts are being examined to enable tape to deliver more native random-like access capability, possibly using new tape geometry. This will play a key strategic role in the evolution of SAN-ready tape libraries acting as a bridge between highspeed front-end access platforms and high-capacity backend archival subsystems.

The growth of digital storage has dictated that the tape industry must now accommodate 7x24x365 operations with higher performance, higher availability, and less human intervention. The tape industry is about to take several giant steps forward and if successful, will continue to define itself as the predominant mass storage platform as the digital age unfolds.

Tape Evolution
First era   1952 - 1987  Manual tape
Second era  1988 - 2000  Automated tape
Third era   2001 +       Intelligent tape