Coda File System

Re: Worries about hoarding

From: Phil Nelson <phil_at_cs.wwu.edu>
Date: Mon, 24 Jan 2000 09:45:04 -0800 (PST)
>I've been thinking about the issue of hoarding.  To my mind, the best
>...
>maintain them in the cache.  The idea is that the cache is memory-mapped,
>so whatever part of it does not fit in memory can be written to swap.
>...
>This guarantees that venus will keep local copies of everything, as
>long as there is enough virtual memory--right?  Please correct me if
>I'm wrong.  I still don't know quite how priorities affect virtual memory
>use.

Although I have not used the windows versions, I'm quite sure that your
cache files are stored on your local disk and do not reside in virtual memory.
The cache manager will have to be run and will take virtual memory, but
the cache files themselves are stored on disk.  You need only to have a
big enough disk to be able to hold all the files needed by the machine's user.
Your initial hoard file seems resonable to me.  Then, just before a 
disconnection, you could simply run "hoard walk". (there should be a good
way to put a gui button up for this.)  The hoard walk will make sure that
any file that is not currently in cache (on the local disk) will be put there.
For most cases, this should be a very short process with very little 
network traffic.  

Starting a new hoard and the corresponding hoard walk just before a disconnection
could cause a longer wait and a lot more network traffic.  It is best to 
have the hoard in place all of the time.

Under UNIX we have to configure the "size" of the cache.  I presume the same
thing happens under windows.  The "standard" size on UNIX is around 100 Megs.
I'm currently running a clint with no problems at 800 Megs.  Startup time of
the cache manager (venus) is a possible problem with larger cache sizes.


-- 
Phil Nelson                    NetBSD: http://www.netbsd.org
e-mail: phil@cs.wwu.edu        Coda: http://www.coda.cs.cmu.edu
http://www.cs.wwu.edu/~phil    
Received on 2000-01-24 12:47:46