VXA technology

Streaming is the technique used to transfer data in linear and helical scan tape drives, which operate by reading an entire track thousands of bytes long in a single pass of a Read head at a fixed tape speed. In order for such drives to operate at peak efficiency, the host must be able to keep the drive's buffer full. A slow host or a busy SCSI bus can cause the tape drive to cease streaming, forcing the drive to reverse, or "backhitch", before resuming the write operation. This not only impairs overall tape-drive performance, but also places additional stress on the media.

Ecrix Corp's VXA technology solves this problem by using variable speed operation, essentially matching the speed of the tape to that of the host, thereby optimising tape drive performance and minimising undue stress on the media. Announced in late 1999, VXA is the first new tape technology in years to take an entirely different approach to data recording. Instead of the long blocks of streaming data used by most tape formats, VXA breaks long strings of data into small data units, then adds a wrapper of verification data, building small data packets - in what the company refers to as Discrete Packet Format (DPF) - before recording it on the media.

DPF allows the packets that comprise a data string to arrive at different times in the data buffer and still be efficiently reassembled into their original string order. Each data packet includes 64 bytes of user data, a synchronisation marker, unique address information, Cyclical Redundancy Check (CRC) code and Error Correction Code (ECC). Each track consists of 387 individual data packets recorded and read through a special buffer array. Two overlapping sets of tracks are written by the two pairs of heads on the drum. The first head in each pair writes the data. The second head performs a Read-After-Write to check the data integrity.

VXA employs four-level error correction which is applied in two phases. First, each packet includes a Reed-Solomon ECC which can correct small errors typically caused by noise or phase shifts. Second, when the packets are collected in the buffer segment, they form an array which employs a three-dimensional Reed-Solomon ECC (X-axis ECC, Y-axis ECC and Diagonal ECC).

The combination of DPF, variable tape speed operation and an overscanning technique which allows data to be read from any physical location on the tape, without having to follow tracks from beginning to end, result in a technology which is fast, reliable and highly immune to media errors. Initial drives operated at 6MBps, provided a maximum compressed storage capacity of 66MB and were shown to be capable of reading tapes that had been frozen and immersed in boiling water!

Robotic applications

There are all kinds of storage device that can be used to back up data, from high density super-floppy disks to DVD drives. But when it come to providing the speed, capacity and reliability needed to protect shared network data, most companies choose one of the automated tape storage applications have emerged in recent years:

• Tape Libraries: The increasing demands being placed on client servers require provisions for off-line data storage with automated access and control. A "tape library" is a high-capacity data storage system for storing, retrieving, reading and writing multiple magnetic tape cartridges. It is essentially two pieces of hardware: the tape drive itself plus a set of robotics. The latter provides the required capacity, picking up tape cartridges from built in storage racks and loading them into the drive as and when required by the backup software. Similarly, the autoloader then removes the cartridges when they're full, storing them until they're next needed. Tape libraries are available for half-inch, QIC, Travan, DAT and 8mm tape cartridges. Smaller units can have several drives for simultaneous reading and writing and may hold from a handful to several hundred cartridges. Large units can have hundreds of drives and hold several thousand cartridges. Tape Library devices are referred to as "near on-line" (or "near-line") because they are not as fast as on-line hard disks.
• Tape Arrays: Tape arrays are based on technology borrowed from disk-based RAID subsystems. The idea isn't new, but until recently arrays did not provide much performance benefit over single drives. They have now become a practical option for midrange storage. Arrays are based on special controllers that can stripe data across multiple drives in parallel, as opposed to the slow sequential access on a single device. Tape array vendors claim that an array with four drives will provide a fourfold performance improvement over a stand-alone drive. For extra fault tolerance, most tape arrays can be configured with a parity drive. The downside is a decrease in overall throughput gains.
• Hierarchical Storage Management: HSM applications are designed to minimise storage costs while optimising performance. They achieve this by combining multiple storage media such as magnetic disk, optical disk, and tape into a single logical unit, and transparently migrating data between media based on access frequency.