Re: Asymmetric networks (ADSL)

The bottleneck is definitely the downstream bandwidth and also the round
trip times. For example for patch subset the number of request bytes is
~2.5% of the total response bytes sent. Given that I don't believe using
the assumption that up bandwidth = down bandwidth will have a material
impact on the simulation results.

One of the limitations of the data set that I derived the network models
from is that it only has information on download bandwidth. So if we do
want to incorporate asymmetric speeds then we'll have to guess at some
approximate ratios between up and down or find an alternate data
source which does have that info.

On Fri, Sep 25, 2020 at 7:44 AM Chris Lilley <chris@w3.org> wrote:

> I notice that all of the network models assume identical download and
> upload speeds.
>
> Asymmetric networks, such as Asymmetric Digital Subscriber Line (ADSL)
> which sends digital signals over ordinary copper phone lines, are
> common. For example, the currently popular ADSL2+ has 24Mbps  downstream
> and up to 1.4Mbps upstream, while the older but still common ADSL has
> 8Mbps  downstream and 448Kbps upstream. These theoretical speeds are
> rarely attained in practice.
>
>
> https://www.increasebroadbandspeed.co.uk/2012/graph-ADSL-speed-versus-distance
>
> (mostly concentrates on downstream speeds)
>
> I *think* that the bottleneck for PFE is downstream throughput, but some
> of the requests for CJK can be quite large and would be affected by
> upstream bandwidth.
>
> I wonder therefore if we should add a Desktop ADSL network model?
>
> (This email sent over ADSL2+)
>
> --
> Chris Lilley
> @svgeesus
> Technical Director @ W3C
> W3C Strategy Team, Core Web Design
> W3C Architecture & Technology Team, Core Web & Media
>
>
>