Re: [Cake] [sch_cake] Merge branch 'exposeinterval' (5541271)

[Dave Taht <dave.taht@gmail.com>]

On Mon, Oct 12, 2015 at 11:08 AM, Sebastian Moeller <moeller0@gmx.de> wrote:
> Hi Kevin,
>
> On Oct 12, 2015, at 10:55 , Kevin Darbyshire-Bryant <notifications@github.com> wrote:
>
>> It's worse/better than that - if you look at cake_set_rate it actually sets the interval to the max of target(5ms) * 8 or the supplied interval parameter, so actually as it currently stands the lowest you can ACTUALLY set interval to is 40ms.

This is incorrect for the DC use case. We should be able to have the
target fall below at least 250us.

>         Good to know, data center people might not appreciate that given that for fq_codel a target value of 500µsec seems to be recommended (now whether fq_codel/cake really are applicable for data centers is another question). But since I have no data center available I would not know how to test this, so others need to chime in.
>         (Note to self, it might make sense to also expose the target-parameter, potentially not in raw time units, but as a percentage of interval, as specified in the codel RFC draft section "4.4.  The target Setpoint,”, see https://datatracker.ietf.org/doc/draft-ietf-aqm-codel/?include_text=1 ).
>
>>
>> I put those limits in purely as an anti hack/tc stupidity/pointer gone wild - defensive programming thing. Take 'em out/tweak/whatever.
>
>         Oh, I am all for sanity and playing it safe, this is why I wondered whether the actual limits will work for the targeted audience. It would be sweet if the initial limits of the upstream commit could survive some time, so sqm-scripts does not need any gymnastics to inform users about changed behavior (so I am asking for purely selfish reasons) ;)
>         Loganaden, since you have experience, how long are your RTTs  from Mauritius to France and both US coasts (as a proxy for “typical” traffic destinations); so we can taylor the limits to allow that.
>
> Best Regards
>         Sebastian
>
>>
>> —
>> Reply to this email directly or view it on GitHub.
>>
>



-- 
Dave Täht
Do you want faster, better, wifi? https://www.patreon.com/dtaht