Quantum LTO-4 (Model B), Руководство пользователя

Страница: 57 / 98

Chapter 4 Theory
Data Compression
LTO-4 (Model B) Half-Height Tape Drive User’s Guide 41
• Drive buffer size
• Host computer input/output (I/O) speed
• Effective disc speeds of the host computer
• Record lengths that the host computer transmits
Data compression algorithms can be tailored to provide maximum
compression for specific types of data. Because varying types of data are
encountered in normal day-to-day operating circumstances, however,
an effective data compression method for a tape drive must serve
various data types. Additionally, the data compression method must
adapt to different data types, automatically providing optimum
handling for all types of data.
Intelligent Data
Compression
The compressed capacity of the tape is maximized through the use of
intelligent data compression. The intelligent data compression hardware
determines the compressibility of each record. If the size of the record is
larger after a compression attempt than the native (uncompressed) size,
then the record is written in its native form.
The intelligent data compression utilizes two compression schemes:
• Scheme-1 is a LZ1-based compression scheme using a history buffer
to achieve data compression.
• Scheme-2 is a pass-through compression scheme designed to pass
uncompressible data through with minimal expansion.
There are three specific requirements for compliance with the LTO
specification.
• The output data stream must be decompressible following LTO rules
to create the input sequence of records and file marks perfectly.
• An LTO compressed data stream may not contain any of the eight
reserved control symbols.
• While control symbols allow switching to Scheme 2, this should
never be used by operational software because this capability is only
for diagnostic and testing purposes.
Software data compression should never be used because the built-in
intelligent data compression of the LTO-4 (Model B) Half-Height Tape
Drive is much more efficient than software data compression.