Re: [ql-developers] SMSQ/E ,HDDs and OpenSMS

[Dave]

On Fri, 8 Feb 2002, Arnould Nazarian wrote:

> So slave blocks make reading files for the first time slow?

This makes perfect sense...

Think of it this way...

When slave blocks were devised, the Microdrive was the storage medium on
the QL. Not *A* storage medium, *THE* storage medium.

Loading data from slow microdrives into slave blocks made no performance
hit (the data was moving more slowly than the caching was).

When floppies and HDs came along, and became viable with the QL, the speed
limit of the caching algorythm became the limiting factor.

I can think of a couple of things that can be addressed to improve/resolve
the situation...

Place a limit on the number of blocks/proportion of free memory that can
be taken by the mechanism.

Have an ability to "defrag" the blocks so old blocks drift to the top, are
vacated, and the memory released.

Set a flag on a file that makes it non-cacheable...

Of course, if we had a whole new filing system...

Dave
ql.spodmail.com