[Bloat] Performance simulation of buffer bloat in routers

[David Täht]

while catching up on the literature and reading the following:

http://staff.science.uva.nl/~delaat/netbuf/bufferbloat_MS-SG.pdf

And "Performance simulation of buffer bloat in routers"

http://staff.science.uva.nl/~delaat/netbuf/bufferbloat_BG-DD.pdf

I note that while otherwise looking very good at first scan, it would be
nice to have several additions to the second paper. I was unable to find
the email addresses of the authors of the second paper, so I'm cc'ing
the first in the hope that they can be contacted.

A) Having a copy of the model's code used would be VERY useful to us...

B) The paper stops the experiments at 500 buffers.

1) the range 1064-1256 is the effective real-world default for ethernet
on most versions of Linux. It would be useful to have a model and
analysis for what happens in the real world (although I suspect they
would have to increase the test duration). I'd like to see theory and
reality line up a little better than they have to date.

2) Effective buffering, with retries, in wireless devices is far, far,
far, in excess of that, at least at present, for many devices. For
reference, going to say, 10,000 buffers, would be good. I look forward
to seeing chaotic effects.

C) The TCP algorithm in use is not documented

D) The distance of the paths is short. Simulating a longer distance path
(say, 100ms) would be interesting.






-- 
Dave Täht

-------------- next part --------------
A non-text attachment was scrubbed...
Name: dave_taht.vcf
Type: text/x-vcard
Size: 204 bytes
Desc: not available
URL: <https://lists.bufferbloat.net/pipermail/bloat/attachments/20111011/8a209110/attachment-0002.vcf>