Adaptive Read Lookahead; Caching Write Data - Seagate Elite 9 Product Manual

Elite 9 Product Manual, Rev. C
5.5.1
5.6
2. The drive prefetches additional logical blocks contiguous to those trans-
ferred in Case A, step 2 above and stores them in the segment. The drive
stops filling the segment when the maximum prefetch value has been
transferred.
During a prefetch, the drive crosses a cylinder boundary to fetch data only
if the Discontinuity (DISC) bit is set to 1 in bit 4 of byte 2 of the Mode Select
parameters page 8h. Default is zero for bit 4 (see the SCSI-2 Interface
Product Manual ).
Each cache segment is actually a self-contained circular buffer whose
length is an integer number of sectors. The drive supports operation with any
integer number of segments 1 to 16. Divide the 960 bytes in the buffer by the
number of segments to get the segment size; default is 3 segments (see the
SCSI-2 Interface Product Manual ). The wrap-around capability of the
individual segments greatly enhances the cache's overall performance,
allowing a wide range of user-selectable configurations including a pure
prefetch strategy.

Adaptive read lookahead

Read lookahead causes the drive to continue reading data from the disc
following a normal read I/O until the read-ahead parameters are satisfied
(prefetch). If subsequent I/O requests can be satisfied from the prefetched
data in the data buffer, there is a significant improvement in performance
since a disc access is eliminated. If subsequent I/O requests cannot be
satisfied from the prefetched data, there is a reduction in performance due
to prefetch overhead.
The adaptive read lookahead feature suspends the prefetch operation if
three subsequent read I/O requests are not satisfied from the prefetched
data. This improves performance because unnecessary prefetches are
avoided. Prefetch is reinstated if an I/O request is sequential to a previous
I/O request.

Caching write data

Write caching is a drive-write operation, which uses a drive's buffer storage
where data to be written to the disc is stored while the drive performs the Write
command.
Write caching is enabled along with read caching. Default is cache enabled.
For write caching, the same buffer space and segmentation is used as set
up for read functions. When a write command is issued, the cache is first
checked to see if any logical blocks to be written are already stored in the
cache from a previous read or write command. If there are, the respective
cache segments are cleared. The new data is cached for subsequent read
commands.
If a 10-byte CDB write command (2Ah) is issued with the Data Page Out
(DPO) bit set to 1, no write data is cached, but the cache segments are still
checked and cleared, if needed, for any logical blocks that are being written
(see the SCSI-2 Interface Product Manual ).
If the number of write data logical blocks exceeds the size of the segment
being written into when the end of the segment is reached, data is written into
the beginning of the same cache segment, overwriting data that was written
there at the beginning of the operation. However, the drive does not overwrite
data that has not yet been written to the disc.
15