Amanda-Users

Re: "netusage" parameter broken?

2003-05-12 13:56:57
Subject: Re: "netusage" parameter broken?
From: Eric Sproul <esproul AT ntelos DOT net>
To: Chris De Young <chd AT chud DOT net>
Date: 12 May 2003 13:52:32 -0400
On Mon, 2003-05-12 at 13:33, Chris De Young wrote:
> On Sun, May 11, 2003 at 07:22:26PM -0700, Stephen Carville wrote:
> > 
> > The netusage value only tell amanda to stop forking dumpers if usage 
> > exceeds the limit.  It does not actually casue amanda to restrict 
> > bandwidth usage.  That is what it is advertised as doing and, AFAIK, 
> > it works correctly.
> 
> Ah, ok... so as long as there is only one client/dumper at any given
> time, that one will go ahead and use as much network bandwidth as it
> can get and there's no way (within Amanda) to throttle it back?  That
> makes sense, although it's rather less useful than I'd hoped.

Chris,
That's it in a nutshell.  The AMANDA chapter of "UNIX Backup and
Recovery" from O'Reilly states:

"The limit is imposed when deciding whether to start a dump, but once a
dump starts, AMANDA lets underlying network components do any
throttling."
http://www.backupcentral.com/amanda-20.html

AMANDA adheres to the UNIX principle of doing one thing, and doing it
well.  It allows flexibility for admins to tweak her performance for
their own needs.  AMANDA is used in many differently-sized
installations, and since network designs are highly site-specific, I
think it's smart that she doesn't try to impose any control over how the
network traffic is treated once it comes out of the client.  

It's the same way with internet protocols. IP is pure transport-- it
doesn't worry about sorting out sessions between applications, it just
worries about getting packets from point A to point B.  It leaves the
session-oriented work (setup/teardown, sequencing, etc.) to TCP.  AMANDA
is an application, and she (correctly, IMO) leaves the transport to a
lower networking layer.

Eric



<Prev in Thread] Current Thread [Next in Thread>