>>> If you watch a real elevator you'll see they too use the elevator
>>> algorithm
>> They stop for passengers going up _or_ down.
> Not elevators I've seen -- they have Up and Down buttons and only
> stop for passengers going in the right direction, or when the
> elevator isn't in use.
I think there's some confusion over the term 'elevator' in this
discussion, so let me clarify the two possible definitions please:
1. US definition: An elevator is a cage suspended on ropes that
moves up or down a vertical shaft depending on the whims of
the people in it. Known as a 'lift' in Europe.
2. European definition: An elevator is a set of stairs that move
up or down (as set) continuously, with passengers stepping on
at one end and off at the other without having to move in
between. They are usually installed in pairs, one going each
way. I don't know what this is called in the USA.
Hopefully, that helps...
On Saturdays in Israel, definition (1) elevators are set to move
continuously, starting at the bottom, going up one floor at a time,
and stopping on each floor until they get to the top, where they then
go down one floor at a time stopping at each floor until they get to
the bottom. This is done because of their beliefs regarding the
Sabbath (Saturday in Israel) being a day when no work can be done,
since one of their senior clergy decided at some time that the spark
of electricity that can be generated when a [lift] button is pressed
constitutes work.
Before anybody complains, I'm NOT picking on them for this belief.
However, their variant of elevator is the one I believe would be the
best one to be used in the Linux kernel for sorting disk accesses.
There have been several claims why it's wrong, but none have been
anywhere near convincing. I'd certainly be interested in seeing one
that is.
I have also had several scientific papers investigating hard disk
access improvement algorithms, two of which I downloaded and printed
out, and both of those concluded that a "two way end-to-end elevator
access method" (as one of them referred to it) was the optimal means
of access. If anybody cares to point out other suchlike papers I can
download and print out, I'll read those as well...
Best wishes from Riley.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/