On Mon, Sep 01, 2003 at 03:25:23PM +0200, Steffen Neumann wrote:
> Mark Phalan <loop_at_netsoc.tcd.ie> writes:
> > Hi,
> > 
> > How well does Coda handle large files (100MB-1GB)?
>
> Basically it does handle them.
> If files are larger than your cache, 
> they won't be cached, even if you accessed
> only parts of the file.
> 
> So they're fetch every time, 
> and written back every time they're changed.

Almost, as any newly created file is 0 bytes initially, the client
allows the application to create the file and open it for writing. Then
the app dumps a couple of 100MB in the file and closes it.

Only at that point we see the real size of the file and if the venus
cache size is set to something smaller than the size of this new file,
venus pretty much says "WTF that's way over the allowed cache size, I
gotta get rid of this one". But it can't because the file is still
dirty.

So after throwing pretty much everything else out of the cache to reduce
the size of the venus cache, the file is sent back to the server (in
some cases we manage to send it back before the cache is pruned), and
finally the large file is thrown out as well because it arrived safely
at the servers and we still have to shrink our cache usage to below the
preset limit.

Now the next time we want to open that file, everything possible is
thrown out of the cache to see if it might fit either way, but when the
file is really large it ofcourse still won't fit and we end up returning
ENOSPC to the application that tries to open the file.

Jan