[GRLUG] NOT LINUX - broadband test

Adam Tauno Williams awilliam at whitemice.org
Sun Jun 27 22:16:27 UTC 2010


On Sun, 2010-06-27 at 15:55 -0400, Michael Mol wrote:
> My argument is that simply taking a start, end and time is
> insufficient when measuring your bandwidth--*especially* when you're
> considering things like how long until a transfer is completed. I'm
> beginning to think that "response curves" like what laptops use to
> predict battery life might soon be applicable to download ETAs. There
> are consistent, accountable variances in throughput.

If TCP is the protocol in question [HTTP?] there is the issue of TCP
window scaling as well;  this can produce exactly [the appearance] of an
initial burst and then a leveling off of 'throughput'.  Tweaking the
window can have measurable effect on transfers.

Also most [all?] discussion of this assume that the constraint on
bandwidth is local.  Throttling of the connection may occur and numerous
points and almost certainly occurs at the remote [host, web site,
content provider, ...].  Throttling traffic from a single source is
pretty common and there is the simple issue of resource constraints on
the remote side of the connection [*assuming* no resource constraints on
the backbone].
-- 
Adam Tauno Williams <awilliam at whitemice.org> LPIC-1, Novell CLA
<http://www.whitemiceconsulting.com>
OpenGroupware, Cyrus IMAPd, Postfix, OpenLDAP, Samba


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.



More information about the grlug mailing list