Network Problem

[d l]

--- Tom 'Needs A Hat' Mitchell <mitch48 at sbcglobal.net> 
からのメッセージ：
> On Tue, Apr 20, 2004 at 08:10:03AM -0600, Eric
> Diamond wrote:
> > Tuesday, April 20, 2004 3:55 AM Andy Green added:
> > > > What units are you using? Mo/sec is not one
> I'm familiar with.
> > > 
> > > Octet, it's used for byte in Francophone
> countries.
> > 
> > Ah! I see said the blnd man...
> > 
> > So that means S�astien was seeing 5 MegaBytes
per
> second vs. 11
> > MegaBytes per second. By my math (and I'm only
> approximating link
> > overhead) that's 50+ MegaBits per second vs. 110+
> MegaBits per second.
> > To me, that sounds like the difference between
> 100Mbps at half and full
> > duplex respectfully.
> 
> 
> If I understand it full duplex does not double the
> download speed.
> But rather upload can progress at the same time as
> download.  This
> can be useful for a file-server where my reads do
> not conflict with
> your writes (on the network wire).
> 
> For download the ACK packets can flow without
> interrupting the
> up/download. Thus it can be possible to approach the
> theoretical 
> bit rate in any one direction.
> 
> Am I confused?
> 
> 
> -- 
> 	T o m  M i t c h e l l 
> 	/dev/null the ultimate in secure storage.
> 
> 
> -- 
> fedora-list mailing list
> fedora-list at redhat.com
> To unsubscribe:
http://www.redhat.com/mailman/listinfo/fedora-list

That depends on flow control method.

With stop&wait flow control max throughput is always 50%
of the max line bit rate, Full duplex helps very little on
this.
protocol like sftp are of this type.

Better flow control methods like sliding windows is very
sensitive to the latency of ACK packets. typical efficency
of s-window is ~60% on half dulplex and 80% on full
duplex.
100% efficency is nvr achived with a single s-window
connection though. http/ftp belong to this cat.


__________________________________________________
Do You Yahoo!?
http://bb.yahoo.co.jp/