[Bloat] Animation

[Jim Gettys]

On 03/01/2011 02:22 AM, Daniel Brooks wrote:
> I saw the video that Richard posted, and I agree that it needs to be
> accompanied by an explanation. Text is the simplest way to go, and
> subtitles can be abused for this purpose, so I converted the video to a
> WebM file and put it up on my webserver at
> http://db48x.net/bufferbloat/. I then used universalsubtitles.org to add
> subtitles to it; You can check them out at
> http://universalsubtitles.org/en/videos/qstl1K7mdGuq/. The subtitles are
> also supposed to show up on db48x.net but for the moment there's a snag.
>
> Anyone can edit these subtitles right in the browser, so I encourage
> anyone who has an idea for improvements to simply do so. Consider the
> existing subtitles to be a first draft. Similarly, if there is any other
> information we can sync to the video in order to produce a better
> presentation then suggest it here in the list and I'll implement
> it. Syncing text, images, whole webpages, and audio to the video is
> practically trivial, so don't be shy. If someone wants to record a
> spoken explanation that would be wonderful.
>
> db48x

The big problem I see (with the original simulation) is that it chooses 
a very unlikely situation to simulate; a number of hops with identical 
throughput, and the queues therefore growing at multiple hops.

Almost always, there will be (for a simple path such as that being 
simulated in this case) a single hop which is the bottleneck, and the 
queues grow on either side of that hop.

Given the original NS2 script, it should be easy to re-run with 
something much more "typical".

Another point we need to get across is that the bottleneck moves and 
therefore where the queues grow: e.g. when your 802.11 bandwidth drops 
below that of your broadband bandwidth and it shifts to your host and 
home router.

Anyone care to take the original ns2 script and regenerate video for the 
typical cases we know happen most often?
			- Jim