[AusNOG] AAB Statement

Mark Smith nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org
Wed Sep 1 19:50:45 EST 2010


On Wed, 1 Sep 2010 13:43:07 +1000
Dmitri Kalintsev <dek735 at gmail.com> wrote:

> Mark,
> 
> Latency has very little effect on streaming video (I assume that this is
> what you mean when referring to the iView). 

No, completely unrelated. People talk about "broadband requirements",
and only seem to ever mention bandwidth as the important attribute.
That'd be true if the only thing we used the Internet for was bulk
data transfer like file transfers, but we don't. For some applications,
latency matters far more than bandwidth, for others, bandwidth matters
far more than latency, for for others, bandwidth and latency matter
generally equally. Jitter is probably something that shouldn't be
forgotten either, although typically it isn't an issue, and
applications adapt to it using things such as playback or jitter
buffers.

My comment about iView was saying that I think a minimum definition
today of broadband in Australia, for the bandwidth attribute, would be
enough to allow the any citizen of Australia to be able to access it.
In other words a ubiquitous minimum of 1.5Mbps or so. That doesn't
necessarily mean to everybody everywhere, but that e.g. they're
commonly within reach of it. So a remote region farmer would have access
to 1.5Mbps at in the home, but not necessarily out in the paddock.

>The only couple of
> latency-sensitive applications of major interest to a typical consumer that
> I could think of off the top of my head would be interactive voice and/or
> video (which needs to have network portion latency to be within ~150-160ms
> one way to perform absolutely top-notch but can easily tolerate up to 200 ms
> one way without user noticing too much) and online gaming (where people
> shoot at each other), where the rule is "the lower the better" without
> actual lower limit.
> 

I think it'd be pretty short sighted to assume that the commonly run
applications of today are going to the same in the future.

The Internet is a dumb packet transport network that makes a best
effort attempt to deliver packets between the end-nodes at it's edge.
The applications also only exist at it's edge (the give away is that you
don't have to upgrade a router when you want to use a new type of
application - you just load up the software on your PC and run it). If
you're going to pick a link layer technology that is going to be widely
deployed and used for general Internet access i.e. you don't know what
types applications are going to be sending traffic over it, then you
don't want to pick a technology that is high bandwidth but has bad
latency, or low bandwidth yet has good latency. As much as you can, you
want to keep the Internet as general purpose as possible so that you
don't create constraints at the link or network layers that the
applications either can't work around or will need to try to (that
sounds like NAT traversal doesn't it ...).

> -- Dmitri
> 
> On Wed, Sep 1, 2010 at 8:02 AM, Mark Smith <
> nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org> wrote:
> 
> > On Tue, 31 Aug 2010 18:16:58 +1000
> > James Spenceley <james at vocus.com.au> wrote:
> >
> > > Noggers,
> > >
> > > As Bev foreshadowed yesterday the 'NBN 3.0: The Alliance for Affordable
> > Broadband" document has now been released.
> > >
> > > If you are interested in adding your name to it please contact one of us.
> > >
> > > Document is available here ...
> > >
> > > http://www.vocus.com.au/media/AAB_Final2.pdf
> > >
> >
> > I'd like to see latency goals mentioned. For a lot of applications, the
> > bandwidth isn't starting to matter as much as the latency (DNS lookup
> > RTTs of e.g. 200ms can make the web look slow).
> >
> > It might also be worth coming with the a definitive definition of what
> > "broadband" is, because you can't really judge what suitable
> > technologies are until you have those parameters. For example, 4G might
> > be able to go up to speeds of 100Mbps, but if it has a link latency of
> > 100ms-200ms (I don't know, I'm making some estimations based on 3G
> > experiences), the amount of bandwidth won't matter to people using
> > interactive or latency sensitive applications.
> >
> > One thought I've had is that people should be able to watch ABC iView
> > as a minimum, as that is our government provided web content source,
> > which means a minimum of 1.5 Mbps for bandwidth, and I'd say something
> > like no more than 50-75ms link latency.
> >
> >
> > > --
> > > James
> > >
> > _______________________________________________
> > AusNOG mailing list
> > AusNOG at lists.ausnog.net
> > http://lists.ausnog.net/mailman/listinfo/ausnog
> >



More information about the AusNOG mailing list