[Csync2] Re: csync2 questions...

[Clifford Wolf]

Hi,

On Mon, Apr 30, 2007 at 02:31:23PM -0700, Drew Bertola wrote:
> [...]
> 
> First, do you recommend csync2 for file systems with a very large number 
> of files ( > 200K files)?  What would the ballpark limit be for number 
> of files to be kept in sync?

It really depends (on i/o throutput, how much of your fs data is in your
block buffers, what are your time constraints, etc..)

csync2 performance scales with something like O(N*log(N)). The larges setup
(in number of files) I do know of is running csync2 with >3 million files.
But sure, in such a setup a check can take already a few minutes..

> What is the algorithm for comparison?  Is it a comparison of the 
> databases stored on each hosts, or do you look at each file as well?

I compare the local database with the local file metadata (size, timestamp,
permissions, etc.). The file content itself is only accessed when actually
doing the sync, not for the check.

> Finally, the customer enjoys tormenting us with a heterogeneous 
> environment.  Have you heard of any success w/ FreeBSD or Solaris? 

no. I've only tried csync2 on linux and on win32 (using cygwin). If there
are any changes needed for FreeBSD or Solaris they should be trivial.

yours,
 - clifford

-- 
It is easier to change the specification to fit the program than vice versa.