On Mon, Apr 04, 2005 at 11:56:38AM -0600, Patrick Walsh wrote:
> 	OK, so there's no penalty for overestimating -- another tidbit that
> should be added to the FAQ and the docs should be updated since I
> thought that the RVM size allocated was kept entirely in memory (not
> just the used portion).

Note, it is valid for files only, for partitions it _does_ take the space
in memory. So, do not use partitions :)

> 	I'm sorry to hit you with so many questions.  Maybe I can take your
> answers and make a little "Optimizing Coda 6.x" document that can
> benefit other people as well.

It would be a very welcome contribution.

> 	Is there any downside to making large (5gb, for example) client caches?

You do not want a too big rvm on the client side, so make sure to explicitely
indicate the max number of files expected in the cache.
You will want to adjust in venus.conf:
--------------------------------
 ...
#
# How many files should the venus cache hold. If this is not specified
# or `0' the default value is calculated as 'cacheblocks / 24'.
# Minimum value is 85.
#
cachefiles=0

#
# How many modification log entries should venus have. If this is not
# specified or `0' the value is calculated as 'cacheblocks / 6'.
# (with the default number of cachefiles this is '4 * cachefiles')
# Minimum value is 256.
#
cml_entries=0

#
# How many hoard database entries should venus have. If this is not
# specified or `0' the value is calculated as 'cacheblocks / 48'.
# (with the default number of cachefiles this is 'cachefiles / 2')
# Minimum value is 42.
#
hoard_entries=0
 ...
--------------------------------
> Is there any tradeoff I should know about when choosing client cache
> size?  Any number I should stay under?

One limitation is that you cannot have more than 1G rvm size.
IIRC if you ask for about 6.5G cache size without adjusting the parameters
above, it will recalculate to >1G RVM which will not work - and is normally
of course unnecessary as it counts for too many cached files, hoarded files
and changes made to the files while disconnected (cml entries).

I run my workstation with an almost 6G cache, more exactly 6000000 K,
with default values as above. It makes venus to grow big as time goes
and more rvm becomes mapped into memory.
Less rvm would probably behave smoother.

Regards,
--
Ivan