Quantum 10K II - Manual (Page 303)

Feature Descriptions
6-2 Quantum Atlas 10K II Ultra 160/m SCSI Hard Disk Drives
• Advanced Ultra3 160/m-LVD, Ultra2-LVD, and Ultra-SE with 68-pin wide
connector or 80-pin SCA-2 connector
• Superior data integrity:
› 352-bit Reed-Solomon quad-burst ECC provides fast on-the-fly correction for
small errors and robust software-based correction for larger errors
› EDC internal protection from SCSI bus to media
Many of these features are described in this chapter.
The drive contains a large, segmented buffer in RAM that is used to maximize the
throughput for sequential read streams. This cache is organized as a number of
separate track line segments. Each track line segment can hold data from one
complete track on the media, temporarily assuming the identity of the track for all
input/output functions.
The buffer is also used to combine or coalesce I/O data on a track-by-track basis,
prior to accessing the disk media. Performance is greatly improved by gathering and
combining the data on a track basis rather than on a command basis. With this
unique system, multiple commands for a particular track are coalesced in the buffer
before accessing the media.
As a result, multiple I/O commands, that address a single track are logically
combined to avoid multiple disk accesses on multiple rotations. Since most reads
and writes are sequential, this greatly increases throughput for most applications.
6.2 PREFETCH
The drive supports prefetch. After the controller has copied an entire media track
into a cache track line segment, a unique algorithm determines whether:
• The most recently used cache track line (track n) satisfies a read request;
• The cache memory also includes another track line (track n-1) corresponding to
the media track preceding the one that was just copied, and;
• There is no other work currently waiting for service by the read/write system.
If all three of these conditions are met, there is a high probability that the host will
call for data from the next media track (track n+1). So, the drive controller
automatically issues a read for that data and substitutes it for the information in
the n-1 track line segment in cache. This process, called prefetch, gives the drive a
very high cache hit rate for sequential reads.
6.3 WRITE-BACK CACHING
The drive supports write-back caching. In this case, writes are considered complete
when the data is loaded into the cache, before it is written to the media. This allows
the host to go on to other tasks, while the drive assumes responsibility for the data
and ensuring that it is written to media as soon as possible. If write-back caching is
employed, it is recommended that an uninterruptible power source (UPS) be
provided for optimal data security.