Open Reel Tapes
In the 1950’s IBM started using half inch wide, ferrous-oxide coated tape, similar to that used for audio recording. Initially, magnetic tape for data storage was wound on 10.5-inch (27 cm) reels. This was the de facto standard for large computer systems and persisted through to the late 1980’s. Tape cartridges and cassettes were available as early as the mid-1970’s and were frequently used with small computer systems. With the introduction of the IBM 3480 cartridge in 1984, large computer systems started to move away from open reel tapes and towards cartridges.
The classification of tape recording methods, generally fall into two categories:
The linear method arranges data in long parallel tracks that run the length of the tape. Multiple tape heads simultaneously write parallel data tracks onto a single medium. This method was used in early tape drives and is the simplest recording method, but also has the lowest data density.
A variation on linear technology is linear serpentine recording, which writes a data tracks on each successive pass along the tape. Each head still writes one track at a time, but after making a pass over the full length of the tape, all heads shift slightly and make another pass in the reverse direction, writing another set of tracks. This procedure is repeated until all tracks have been read or written. By using the linear serpentine method, the tape medium can have many more tracks than read/write heads. The use of the word serpentine is in many cases a misnomer, as many tape formats such as the DLT, record the first track in the centre, moving the read/write head up and down for each successive track, in an outward spiral.
Scanning recording methods write short dense tracks across the width of the tape medium, rather than along the length of it. Tape heads are placed on a rotating drum or disk which rotates rapidly, while the relatively slow moving tape passes over it.
An early method used to get a higher data rate than the prevailing linear method was transverse scan. In this method a spinning disk, with the tape heads embedded in the outer edge, is placed perpendicular to the path of the tape. Another early method was arcuate scan. In this method, the heads are on the face of a spinning disk which is laid flat against the tape. The path the tape heads makes is an arc.
Helical scan recording writes short dense tracks in a diagonal manner. This recording method is used by virtually all videotape systems and several tape data formats. Helical scan recording is able to immediately read the track which has been written, to determine if it was recorded correctly, and if necessary the data will be written again up to a pre-determined error level, beyond which a media flaw is reported. Early helical scan recordings could suffer alignment issues, with the guide pins moving the tracking out of alignment. This could lead to a tape not being able to be read using another drive, and in extreme cases, the new track could overlap the previous one to such an extent as to render it unreadable without special alignment tools. Later drives featured automatic tracking capabilities, making this phenomenon a rare occurrence.
In a typical format, data is written to tape in blocks with inter-block gaps between them, and each block is written in a single operation with the tape running continuously during the write. However, since the rate at which data is written or read to and from the tape drive is not deterministic, a tape drive usually has to cope with a difference between the rate at which data is transferred on and off the tape and the rate at which data is supplied or demanded by its host.
Various methods have been used alone and in combination to cope with this difference. The tape drive can be stopped, backed up, and restarted (known as shoe-shining, because of increased wear of both the media and the head). A large memory buffer can be used to queue the data.
Modern tape drives use speed matching, where the drive can dynamically decrease physical tape speed, by as much as 50% to avoid shoe-shining. Another technique that has been implemented, is to re-write data, or even dummy blocks, but this reduces the capacity of the tape.
The size of the inter-block gap is constant, while the size of the data block is based on the number of bytes in the block. Thus a given length tape reel can hold much less data when written with smaller block sizes than with larger.
Tapes are characterised as sequential access devices, although it is possible to seek to any given tape block as required. To facilitate access to any given file within a backup, many formats will store a copy of the catalogue on tape, which can also be stored on disk. This allows a selective restore, to be made, without the need to read all data blocks on the tape, while searching for the required files. The earliest backup formats did not however store any catalogue, and seeking to a particular file required reading every block until it was located.
Many tapes can be partitioned, allowing data to be stored separate from the main data, although in practice this is rarely done. Some backup formats will store a copy of the catalogue at the start of the process, while some will store a copy at the end, usually separated by a tape file mark, which can be searched for quickly. Some formats keep a catalogue stored on disk, sometimes forgoing keeping one on the tape.
Many modern tape backup formats allow data from multiple sources to be written to tape at the same time. This helps to increase the data throughput when recording, but from a data recovery standpoint, adds many complexities, but nothing insurmountable.