[Cake] Diffserv LLT mode

[Cake] Diffserv LLT mode

[Toke Høiland-Jørgensen]

Is anyone actually using the LLT diffserv setting? The draft describing
it seems to have expired ages ago:

https://datatracker.ietf.org/doc/draft-you-tsvwg-latency-loss-tradeoff/

-Toke

Re: [Cake] Diffserv LLT mode

[Toke Høiland-Jørgensen]

Toke Høiland-Jørgensen <toke@toke.dk> writes:

> Is anyone actually using the LLT diffserv setting? The draft describing
> it seems to have expired ages ago:
>
> https://datatracker.ietf.org/doc/draft-you-tsvwg-latency-loss-tradeoff/

Also, adding to the diffserv thing, right now there seems to be a hidden
ABI where the different diffserv modes are hard-coded in tc. That is not
good for upstreaming, that should go into the header; I'll clean that
up, but thought I'd remove the LLT mode while I'm at it if (as I
suspect) no one actually needs it...

-Toke

Re: [Cake] Diffserv LLT mode

[Dave Taht]

kill it.

On Fri, Apr 20, 2018 at 12:39 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> Toke Høiland-Jørgensen <toke@toke.dk> writes:
>
>> Is anyone actually using the LLT diffserv setting? The draft describing
>> it seems to have expired ages ago:
>>
>> https://datatracker.ietf.org/doc/draft-you-tsvwg-latency-loss-tradeoff/
>
> Also, adding to the diffserv thing, right now there seems to be a hidden
> ABI where the different diffserv modes are hard-coded in tc. That is not
> good for upstreaming, that should go into the header; I'll clean that
> up, but thought I'd remove the LLT mode while I'm at it if (as I
> suspect) no one actually needs it...
>
> -Toke
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake



-- 

Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619

Re: [Cake] Diffserv LLT mode

[Pete Heist]

> On Apr 19, 2018, at 4:50 PM, Toke Høiland-Jørgensen <toke@toke.dk> wrote:
> 
> Is anyone actually using the LLT diffserv setting?

I re-tested diffserv-llt quickly using the current head of cake and tc-adv and am confused by something:

http://drhleny.cz/bufferbloat/cake/round5/diffserv_rrul_eg_cake_ds-llt_50mbit/index.html

When I look at the qdisc stats for cake, there are 5 tins: Low Loss, Best Effort, Low Delay, Bulk and Net Control. Shouldn’t the Low Delay tin have its target and interval reduced to something below 5ms and 100ms? And the bulk tin has a target and interval of 5.8ms and 100.8ms(?)

In short, I forget how diffserv-llt was supposed to work, regardless of whether or not it’s on the chopping block. No need for a history lesson if it will be removed… :)

Improved intra-flow latency could be useful for HTTP/2 or other things, but I don’t see how the current llt mode helps most ordinary people with that, when it would only apply to flows with certain diffserv markings, and doesn’t seem to adjust the target and interval anyway…

Re: [Cake] Diffserv LLT mode

[Y]

Hi.

mine stat doen't  inclued more.
sad.

yutaka

pi@raspberrypi:~ $ tc -s -d qdisc
qdisc noqueue 0: dev lo root refcnt 2 
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) 
 backlog 0b 0p requeues 0
qdisc cake 8017: dev eth0 root refcnt 2 bandwidth 810Kbit diffserv3
srchost rtt 100.0ms raw atm overhead -4 Sent 42114643 bytes 197572 pkt
(dropped 232, overlimits 73726 requeues 0) backlog 0b 0p requeues 0

qdisc ingress ffff: dev eth0 parent ffff:fff1 ---------------- 
 Sent 192984793 bytes 176545 pkt (dropped 0, overlimits 0 requeues 0) 
 backlog 0b 0p requeues 0
qdisc cake 8016: dev ifb0 root refcnt 2 bandwidth 11500Kbit diffserv3
dsthost rtt 100.0ms raw atm overhead -4 Sent 195434409 bytes 176530 pkt
(dropped 14, overlimits 107096 requeues 0) backlog 0b 0p requeues 0


On Sat, 21 Apr 2018 10:44:19 +0200
Pete Heist <pete@eventide.io> wrote:

> > On Apr 19, 2018, at 4:50 PM, Toke Høiland-Jørgensen <toke@toke.dk>
> > wrote:
> > 
> > Is anyone actually using the LLT diffserv setting?  
> 
> I re-tested diffserv-llt quickly using the current head of cake and
> tc-adv and am confused by something:
> 
> http://drhleny.cz/bufferbloat/cake/round5/diffserv_rrul_eg_cake_ds-llt_50mbit/index.html
> 
> When I look at the qdisc stats for cake, there are 5 tins: Low Loss,
> Best Effort, Low Delay, Bulk and Net Control. Shouldn’t the Low Delay
> tin have its target and interval reduced to something below 5ms and
> 100ms? And the bulk tin has a target and interval of 5.8ms and
> 100.8ms(?)
> 
> In short, I forget how diffserv-llt was supposed to work, regardless
> of whether or not it’s on the chopping block. No need for a history
> lesson if it will be removed… :)
> 
> Improved intra-flow latency could be useful for HTTP/2 or other
> things, but I don’t see how the current llt mode helps most ordinary
> people with that, when it would only apply to flows with certain
> diffserv markings, and doesn’t seem to adjust the target and interval
> anyway… _______________________________________________ Cake mailing
> list Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] Diffserv LLT mode

[Jonathan Morton]

>> Is anyone actually using the LLT diffserv setting?
> 
> I re-tested diffserv-llt quickly using the current head of cake and tc-adv and am confused by something:
> 
> http://drhleny.cz/bufferbloat/cake/round5/diffserv_rrul_eg_cake_ds-llt_50mbit/index.html
> 
> When I look at the qdisc stats for cake, there are 5 tins: Low Loss, Best Effort, Low Delay, Bulk and Net Control. Shouldn’t the Low Delay tin have its target and interval reduced to something below 5ms and 100ms? And the bulk tin has a target and interval of 5.8ms and 100.8ms(?)
> 
> In short, I forget how diffserv-llt was supposed to work, regardless of whether or not it’s on the chopping block. No need for a history lesson if it will be removed… :)
> 
> Improved intra-flow latency could be useful for HTTP/2 or other things, but I don’t see how the current llt mode helps most ordinary people with that, when it would only apply to flows with certain diffserv markings, and doesn’t seem to adjust the target and interval anyway…

Well, at some point the intended functionality seems to have got corrupted, though it could be put back in without difficulty.  But I'm certainly not among LLT's userbase; I put the mode in mostly as a demonstration that Cake could do it if required.  If the originators are no longer pursuing the proposal, then the mode has no reason to exist any more.

The bulk tin often gets its parameters adjusted, by virtue of the fact that its threshold is a small fraction of the total.  It happens in the other modes as well.  Don't take that as significant.

 - Jonathan Morton

Re: [Cake] Diffserv LLT mode

[Pete Heist]

[-- Attachment #1: Type: text/plain, Size: 1060 bytes --]


> On Apr 21, 2018, at 6:44 PM, Jonathan Morton <chromatix99@gmail.com> wrote:
>> 
>> Improved intra-flow latency could be useful for HTTP/2 or other things, but I don’t see how the current llt mode helps most ordinary people with that, when it would only apply to flows with certain diffserv markings, and doesn’t seem to adjust the target and interval anyway…
> 
> Well, at some point the intended functionality seems to have got corrupted, though it could be put back in without difficulty.  But I'm certainly not among LLT's userbase; I put the mode in mostly as a demonstration that Cake could do it if required.  If the originators are no longer pursuing the proposal, then the mode has no reason to exist any more.
> 
> The bulk tin often gets its parameters adjusted, by virtue of the fact that its threshold is a small fraction of the total.  It happens in the other modes as well.  Don't take that as significant.

I see. Well, since it’s not working anyway and we haven’t heard anything about it, I’m also for carting it away...

[-- Attachment #2: Type: text/html, Size: 3843 bytes --]

Re: [Cake] Diffserv LLT mode

[Toke Høiland-Jørgensen]

Pete Heist <pete@eventide.io> writes:

>> On Apr 21, 2018, at 6:44 PM, Jonathan Morton <chromatix99@gmail.com> wrote:
>>> 
>>> Improved intra-flow latency could be useful for HTTP/2 or other things, but I don’t see how the current llt mode helps most ordinary people with that, when it would only apply to flows with certain diffserv markings, and doesn’t seem to adjust the target and interval anyway…
>> 
>> Well, at some point the intended functionality seems to have got corrupted, though it could be put back in without difficulty.  But I'm certainly not among LLT's userbase; I put the mode in mostly as a demonstration that Cake could do it if required.  If the originators are no longer pursuing the proposal, then the mode has no reason to exist any more.
>> 
>> The bulk tin often gets its parameters adjusted, by virtue of the fact that its threshold is a small fraction of the total.  It happens in the other modes as well.  Don't take that as significant.
>
> I see. Well, since it’s not working anyway and we haven’t heard
> anything about it, I’m also for carting it away...

Already nuked it from the cobalt branch :)

-Toke

Re: [Cake] Diffserv LLT mode

[Toke Høiland-Jørgensen]

Y via Cake <cake@lists.bufferbloat.net> writes:

> From: Y <intruder_tkyf@yahoo.fr>
> Subject: Re: [Cake] Diffserv LLT mode
> To: cake@lists.bufferbloat.net
> Date: Sat, 21 Apr 2018 21:40:48 +0900
>
> Hi.
>
> mine stat doen't  inclued more.
> sad.

Did you upgrade the kernel module without updating tc, or vice versa?

-Toke

Re: [Cake] Diffserv LLT mode

[Y]

Hi.
I did and reboot.
but this.

yutaka

pi@raspberrypi:~ $ tc -d -s qdisc
qdisc noqueue 0: dev lo root refcnt 2 
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0) 
 backlog 0b 0p requeues 0
qdisc cake 8002: dev eth0 root refcnt 2 bandwidth 810Kbit (?diffserv?)
srchost testflags 502044 rtt 100.0ms raw atm overhead -4 Sent 148612
bytes 1147 pkt (dropped 1, overlimits 94 requeues 0) backlog 0b 0p
requeues 0

qdisc ingress ffff: dev eth0 parent ffff:fff1 ---------------- 
 Sent 304061 bytes 1159 pkt (dropped 0, overlimits 0 requeues 0) 
 backlog 0b 0p requeues 0
qdisc cake 8001: dev ifb0 root refcnt 2 bandwidth 11500Kbit
(?diffserv?) dsthost testflags 502044 rtt 100.0ms raw atm overhead -4
Sent 320245 bytes 1159 pkt (dropped 0, overlimits 166 requeues 0)
backlog 0b 0p requeues 0

Re: [Cake] Diffserv LLT mode

[Jonathan Morton]

> qdisc cake 8002: dev eth0 root refcnt 2 bandwidth 810Kbit (?diffserv?)

This suggests that the kernel is in a Diffserv mode that the userspace tc version doesn't understand.  Since Toke has just been fiddling with that, you should ensure you update and recompile *both* the kernel module and the iproute2 package.

 - Jonathan Morton

Re: [Cake] Diffserv LLT mode

[Y]

Hi.

I cant. then I adjust fq_codel.
ty :)
yutaka

On Mon, 23 Apr 2018 07:39:12 +0300
Jonathan Morton <chromatix99@gmail.com> wrote:

> > qdisc cake 8002: dev eth0 root refcnt 2 bandwidth 810Kbit
> > (?diffserv?)  
> 
> This suggests that the kernel is in a Diffserv mode that the
> userspace tc version doesn't understand.  Since Toke has just been
> fiddling with that, you should ensure you update and recompile *both*
> the kernel module and the iproute2 package.
> 
>  - Jonathan Morton
>