* [LARTC] tail and head drop qdiscs
@ 2005-01-19 21:42 vincent perrier
2005-01-25 12:32 ` Andy Furniss
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: vincent perrier @ 2005-01-19 21:42 UTC (permalink / raw)
To: lartc
I think that there are no qdiscs that permit to drop the oldest
frame of a queue when this queue is full, but I would like to
be wrong:
bfifo drops arriving frames when the max queue length is reached.
red also drops arriving frames in a more elaborate fashion, with
a drop probability that increases above a limit and becomes
a drop certitude when the max queue length is reached.
sfq drops "fairly", it piles frames with same (ip source, ip dest,
port source and port dest) and drops a frame from the
biggest pile to favour the smaller flows.
But I cannot find the qdisc that reacts like a bfifo or red qdisc
but drops the oldest message of the queue when a drop is needed.
Even better: a qdisc with an autodestruct algorithm for each
frame in the queue that would garantee the "freshness" of
each frame coming out.
If you know of such qdiscs, please share.
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [LARTC] tail and head drop qdiscs
2005-01-19 21:42 [LARTC] tail and head drop qdiscs vincent perrier
@ 2005-01-25 12:32 ` Andy Furniss
2005-01-25 19:41 ` vincent perrier
2005-01-27 13:21 ` Andy Furniss
2 siblings, 0 replies; 4+ messages in thread
From: Andy Furniss @ 2005-01-25 12:32 UTC (permalink / raw)
To: lartc
vincent perrier wrote:
> I think that there are no qdiscs that permit to drop the oldest
> frame of a queue when this queue is full, but I would like to
> be wrong:
>
> bfifo drops arriving frames when the max queue length is reached.
>
> red also drops arriving frames in a more elaborate fashion, with
> a drop probability that increases above a limit and becomes
> a drop certitude when the max queue length is reached.
>
> sfq drops "fairly", it piles frames with same (ip source, ip dest,
> port source and port dest) and drops a frame from the
> biggest pile to favour the smaller flows.
>
> But I cannot find the qdisc that reacts like a bfifo or red qdisc
> but drops the oldest message of the queue when a drop is needed.
>
> Even better: a qdisc with an autodestruct algorithm for each
> frame in the queue that would garantee the "freshness" of
> each frame coming out.
>
> If you know of such qdiscs, please share.
>
What sort of thing do you want this for?
I made a version of esfq head drop for use in specific circumstances.
The only other thing I thought head drop generally could be usefull for
is a game server, but then maybe a policer would be better than a queue
WRT latency.
Andy.
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [LARTC] tail and head drop qdiscs
2005-01-19 21:42 [LARTC] tail and head drop qdiscs vincent perrier
2005-01-25 12:32 ` Andy Furniss
@ 2005-01-25 19:41 ` vincent perrier
2005-01-27 13:21 ` Andy Furniss
2 siblings, 0 replies; 4+ messages in thread
From: vincent perrier @ 2005-01-25 19:41 UTC (permalink / raw)
To: lartc
> vincent perrier wrote:
> > I think that there are no qdiscs that permit to drop the oldest
> > frame of a queue when this queue is full ...
> What sort of thing do you want this for?
> ...
> Andy.
>
Very low rate (satellite links for example) and strict priority can
lead to messages waiting for a chance to be transmitted and getting
old in low prio queues as higher priority messages take all the
bandwidth.
Vincent.
> _______________________________________________
> LARTC mailing list / LARTC@mailman.ds9a.nl
> http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
>
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [LARTC] tail and head drop qdiscs
2005-01-19 21:42 [LARTC] tail and head drop qdiscs vincent perrier
2005-01-25 12:32 ` Andy Furniss
2005-01-25 19:41 ` vincent perrier
@ 2005-01-27 13:21 ` Andy Furniss
2 siblings, 0 replies; 4+ messages in thread
From: Andy Furniss @ 2005-01-27 13:21 UTC (permalink / raw)
To: lartc
vincent perrier wrote:
>>vincent perrier wrote:
>>
>>>I think that there are no qdiscs that permit to drop the oldest
>>>frame of a queue when this queue is full ...
>
>
>>What sort of thing do you want this for?
>>...
>>Andy.
>>
>
>
> Very low rate (satellite links for example) and strict priority can
> lead to messages waiting for a chance to be transmitted and getting
> old in low prio queues as higher priority messages take all the
> bandwidth.
> Vincent.
Ahh - if the messages established tcp then I don't think it will make
any difference.
If the messages are udp then I suppose it could help a bit - depends
what you mean by messages.
Andy.
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2005-01-27 13:21 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-01-19 21:42 [LARTC] tail and head drop qdiscs vincent perrier
2005-01-25 12:32 ` Andy Furniss
2005-01-25 19:41 ` vincent perrier
2005-01-27 13:21 ` Andy Furniss
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.