Yup: author makes a classic, and I fear common, wrong assumption about
tradeoffs in queuing systems.
Latency is minimized /throughout/ the operating range if a queue is not
allowed to form. TCP consciously sees a queue as congestion and avoids
having it.
Bufferboat causes queuing, and degrades both (low) latency and
throughput, trying to drive the link toward congestive collapse. The
worst of all possible worlds!
In this case it also tries to drive Dave Taht's circulation system into
collapse, which is doubly ungood.
--dave
On 10/30/2014 09:49 AM, Dave Taht wrote:
I didn't get past the first sentence.
"The question boils down to quantify buffer sizes and yet achieve 100%
utilization on links with maximum
throughput at a feasible cost. "
My goal has always been to have minimal induced latency and reasonable
utilization, and also to keep my blood pressure low. Reading further
strikes me as damaging to both goals.
On Thu, Oct 30, 2014 at 5:12 AM, Fred Baker (fred) <[email protected]> wrote:
Folks from AQM may be interested to comment on
draft-ksubram-lmap-router-buffer-sizes on the lmap list.
_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm
--
David Collier-Brown, | Always do right. This will gratify
System Programmer and Author | some people and astonish the rest
[email protected] | -- Mark Twain
_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm
