Hello Chris,

On Sat, May 01, 2004 at 12:12:27PM -0400, Chris Woodfield wrote:
> Coda seems to be a perfect solution here, at least as soon as it's 
> working and stable under MacOS X...my question is whether or not Coda 
> supports a kind of "full hoard mode" where I can run a Coda server on 
> the Linux machine, and each client workstation stores all files 
> locally, syncing file changes to the server as they occur.

theoretically you can hoard as much as you like - but the hoard
implementation does not scale very well.
Your venus will end up using a lot of cpu and virtual memory, and most
important, keeping accessing the memory, not letting it to be swapped out
for a long time. A lot of physical RAM will help, of course.

I have tested a setup with about 6G cache which was populated
by recursive find /coda/myrealm -type f -exec cp {} /dev/null or alike.
Then the client had all the files locally (the whole realm contains about 5G)
and was pretty much usable. Of course, files not accessed from the client
are not updated in the cache even when they are changed from another client.

A half-hearted try to setup a "hoard all" client failed because
of hoarding being too resource-intensive.

Another limitation I have observed was that Venus cache size is
limited by about 8G for Linux (due to 1G RVM size limitation).
In some circumstances it is even as little as 768M RVM, yielding
about 6G max cache size. I suppose MacOS has a similar limitation
on virtual memory mapping.

Yet nonexistent "next generation" recoverable memory model will solve it,
but for the moment count with 8G max available cache size.

> From the FAQ, it's unclear whether or not a client that is hoarding 
> files continues to sync them to the server if it is connected, or if 
> the sync only takes place when hoard mode is activated and deactivated.

It is keeping files in sync continuously while connected.

Regards,
--
Ivan