Subject: Re: HPT366 (UDMA/66 patch)
To: Roger Brooks <R.S.Brooks@liverpool.ac.uk>
From: Hal Murray <firstname.lastname@example.org>
Date: 03/10/2000 13:58:28
> Start reading from mode 4 drive - iostat shows 25 MB/s
> Start reading from mode 2 drive - mode 4 drive slows down, so both
> are at 9 MB/sec
> Finish reading from mode 2 drive - mode 4 drive goes back to 25 MB/sec.
Here is a crazy suggestion that might help somebody work around that
quirk. Assume the disk has a read-ahead buffer. If the data is
in the buffer, a read will be fast. If it's not, the read will take
a while for the arm to get to the right cylinder and the head to
rotate over the desired bits.
So time the read. If it was fast, process the next read on the same
drive if there is one queued up. If it was slow, switch to the other
drive, if there is work queued up there.
That's not the optimal way to do things, but it might be simple to
implement and get a lot of performance improvement.