Coda File System

Re: bandwidth measure and disconnections

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 1 Jul 2004 15:36:27 -0400
On Thu, Jul 01, 2004 at 10:56:49AM -0300, Juan Carlos Schroeder wrote:
> Ok, thanks Jan. I think I've understood the why "write-back cache" exists.
> Is this that you're saying written in any document? Does exist something
> like a FAQ of Coda?

Not sure if it is written down anywhere. In any case it never worked
right so it will be removed and there is enough undocumented stuff
around to not bother about failed experiments or obsolete features.

> a) The doubt I still have is about bandwidth measurement. I'm having some
> problems with this. Is this working ok?

It is only an estimate, it doesn't try to measure the actual link
capacity, but only estimate available bandwidth. This is difficult
because network usage is bursty.

> b) Also, I would like to know how Coda would work with a lower "limit"
> ("threshold") where to decide to be "fully-connected" or
> "write-disconnected". In the "Coda File System User and System

There is a venus commandline option '-weakthresh' which takes as
argument the threshold at which to go to weak mode in bytes/sec. 'venus
-weakthresh 10240' should set it to 10KB/s. There is also 'cfs strong'
which tells venus to ignore the trigger that occurs whenever we cross
the threshold in either direction. So that could be used to force venus
to work only connected or disconnected, or to stay (write-)disconnected.

Jan
Received on 2004-07-01 15:42:57