Thread (18 messages) flat view 18 messages, 5 authors, 2012-01-13

Re: [PATCH net-next] net_sched: red: split red_parms into parms and vars

From: Dave Taht <hidden>
Date: 2012-01-06 08:31:47

On Fri, Jan 6, 2012 at 6:47 AM, Eric Dumazet [off-list ref] wrote:
Le jeudi 05 janvier 2012 à 14:39 +0100, Eric Dumazet a écrit :
quoted
Problems of RED are :

1) the 'average' on highly variable trafic : It means nothing
interesting at all.

2) the idle time management requires to specify the link bandwidth, as a
fixed value. This is a problem.
A third problem is that ECN marking is performed on the packet we add to
the tail of the queue. It's very late...

I am currently testing my SFQRED and I try to add ECN in the first
packet in queue so that TCP can react faster.
If you are in particular looking for stuff worth marking, vs dropping...
you could even look deeper in the queue (group) than the first packet.

I am reminded somewhat of how david mills handled the first
NSFnet congestion collapse on the fuzzball.

http://osdir.com/ml/culture.internet.history/2004-12/msg00003.html

"When a new packet arrived and no buffer space was available, the
output queues were scanned looking for the biggest elephant (total
byte count on all queues from the same IP address) and killed its
biggest packet. Gunshots continued until either the arriving packet
got shot or there was enough room to save it. It all worked
gangbusters and the poor ftpers never found out. "


-- 
Dave Täht
SKYPE: davetaht
US Tel: 1-239-829-5608
FR Tel: 0638645374
http://www.bufferbloat.net
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help