Coda File System

Re: R: R: R: coda + reiserfs

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 19 Jul 2001 12:14:59 -0400
On Thu, Jul 19, 2001 at 06:02:55PM +0200, Andrea Cerrito wrote:
> > I compare Coda more with squid than with NFS, and hope that at some
> > point it will be normal to run with a 6GB venus.cache.
> 
> I don't understand: if I cannot make a server with > 9gb of data, how can I
> have a 6gb cache?
> I've set my cache to 20mb, for mail servers and webservers. Do you think
> I've to increase it? My codafs are all of 9gb of data.

A bigger client cache improves caching, reduces network and server load
and allows for having most interesting things cached during disconnected
operation. Our webserver's Coda client has a 100MB cache, my desktop has
250MB.

A larger cache also has disadvantages, longer startup time, and more
data to revalidate after a disconnection. Typically I try to get the
cache-size between 1x and 2x the 'active working set'. But caches larger
than about 200MB are sure to tickle a few problems as some cache-wide
operations don't scale nicely.

> And if it so, will be sufficient stopping services, stopping venus, modifing
> the conf, and restart all?

No, you would require a venus reinit, the larger cache can hold more
files. As a result we need more RVM to store the file metadata, which is
only allocated when the client is (re)initialized.

> > We could modify some benchmark app that can measure the difference so
> > that we can decide on the 'optimal' layout before starting the client.
> > f.i. modify the postmark benchmark to compare a run in a hierarchical
> > tree vs. a flat directory layout.
> 
> Not a bad idea... at least, we can shut down venus, modify the option, and
> restart it with --init to be sure the all will be in the right place.

That reminds me, I still need to add something to venus to remove old
containerfiles out of the venus.cache tree when we're reinitializing.

Jan
Received on 2001-07-19 12:15:02