Speed problem

[uwp at dicke-aersche.de]

On Tue, 12 Nov 2002, Wayne Davison wrote:

> On Tue, Nov 12, 2002 at 04:32:31PM +0100, uwp at dicke-aersche.de wrote:
> > I'd call it a bug.
>
> No, it's not a bug.  It's the heart of the rsync algorithm at work.

Hm...

> Rsync trades CPU and local file I/O for network I/O in order to reduce
> the amount of data that is transferred over the network.  Your diagnosis
> has just shown that when the network I/O dips, rsync has traded it for
> local I/O (grabbing matching blocks from the current file instead of
> asking for it to be sent over the network).  For really large files that

And why it tries to get 100% CPU even though there's nothing to do ?
Maybe I don't understand your point here.

> The only alternative is to use the --whole-file option -- this option
> turns off the rsync algorithm and just sends all the changed files over
> the net completely (like an scp copy, but for changed files).  This

Ok, that I never tried because I thought the --partial option should
have been the fastest method because lots of data is still on the other
side if an error has occured before. I'll try that tomorrow, maybe
we'll come closer to the scp.

> should only be used if you have a really fast network connection OR if
> you don't want to trade the CPU and local I/O for network I/O.

Both servers are extremely fast, the slowest parts of them maybe are the
CPUs...

Anyway, maybe we're using it not the way it's been invented, but it's the
only software I've known upto today that has every feature I need to do
complete 1:1 copies of big systems. And even though I don't understand
everything of the concept of rsync it's a fantastic software.
Thank you for this fine tool !

Mermgfurt,
		Udo
-- 
      Udo Wolter <-> uwp at dicke-aersche.de        |  /"\
  !!! Free Music Video !!! All Linux made !!!    |  \ / ASCII RIBBON CAMPAIGN
http://www.dicke-aersche.de/chapterx/video.html  |   X    AGAINST HTML MAIL
   !!! First Music Video made with Linux !!!     |  / \