john@jwt.UUCP (John Temples) (12/07/89)
I have a WD1006V-MM2 compatible hard disk controller which has track caching. Does anyone know how the track caching feature works in reading and writing? I installed one of these controllers on my UNIX system, and the data read rate doubled, yet the write rate went down about 25%. This is in a slow 286 system with the drive formatted 3:1 interleave. I then tried reformatting to 1:1 interleave. Read performance was still about double, yet write performance was down to about 50% of its original value. Does this mean that a track is still read in in a single disk revolution regardless of the interleave, and that writes are not cached? -- John W. Temples -- john@jwt.UUCP (uunet!jwt!john)
cs4g6ag@maccs.dcss.mcmaster.ca (Stephen M. Dunn) (12/10/89)
In article <767@jwt.UUCP> john@jwt.UUCP (John Temples) writes:
$ [...] Does this mean that a track is still read
$in in a single disk revolution regardless of the interleave, and that
$writes are not cached?
For the track reading, the limiting factor that makes interleave
adjustment necessary for optimum performance is not the interface between
the bard disk and the controller, but rather that between the controller
and the processor's memory. The disk controller should be able to handle
all seventeen sectors at once and store them in memory, so interleave
shouldn't make a difference to read times.
As for writes ... it sounds like the controller uses a write-through cache,
which means that data is written to the disk as soon as you send it, rather
than when it's convenient. This would still be affected by interleave, as
the controller has to wait until the correct sector is passing under the
head before it can write. Write-through caches are more reliable in
situations where there may be a power failure at any time, since it
ensure that your data is written to the disk immediately (well, pretty
much). However, it would improve write times significantly to queue
disk write requests on the controller board.
There is still one thing, though, that is odd - the decrease in disk
write performance even without changing interleave. I can't think of
any reason why that would happen.
--
Stephen M. Dunn cs4g6ag@maccs.dcss.mcmaster.ca
<std_disclaimer.h> = "\nI'm only an undergraduate!!!\n";
****************************************************************************
If it's true that love is only a game//Well, then I can play pretend