Re: Filing system again...

[Tommy Thorn <tthorn@daimi.aau.dk> writes:]

> > ... Both srv/tmpfs and srv/vstafs have
> > unique cache handling because of this.
>Doesn't prevent a dynamic buffer cache? A big loss in my view.

If I'm parsing this correctly, you're saying "But how do we do a dynamic
buffer cache?"

I'm open to suggestions. I will point out that the size of the cache
within each filesystem is much bigger than I would have made it in a
traditional kernel. Since the buffers are in virtual memory, it will
in essence be grown or shrunk in response to the overall system memory
availability. It gets this feature without *any* server code complication
at all.

Arguably, the reason you limit the size of a virtual memory-based cache
is simply to bound the consumption of swap space. Other thoughts are
welcome; the buffer cache story is by no means complete!

                                                Andy
Received on Tue Dec 21 08:55:58 1993