[Rpm] [Bloat] [Cake] The most wonderful video ever about bufferbloat

Sebastian Moeller moeller0 at gmx.de
Mon Oct 10 16:23:16 EDT 2022


Hi Erik,


> On Oct 10, 2022, at 13:46, Taraldsen Erik <erik.taraldsen at telenor.no> wrote:
> 
> 
> 
> On 10/10/2022, 11:41, "Bloat on behalf of Sebastian Moeller via Bloat" <bloat-bounces at lists.bufferbloat.net on behalf of bloat at lists.bufferbloat.net> wrote:
> 
>    	[SM] Cool, if I might ask what fraction of the sync are you setting the traffic shaper for and are you doing fine grained overhead accounting (or simply fold that into a grand "de-rating"-factor)?
> 
> We ended up just using a fraction.  

	[SM] Fair enough, for ATM/AAL5 that is challenging but for VDSL2/PTM that seems workable...

> Can't remember the exact fraction, but we were not conservative.  It was hard to push through this change so leaving any bw on the table was sacrilegious to a lot of people.

	[SM] Tricky... e.g. vectoring enabled CPE can be instructed by the DSLAM to send error samples in-band with the data, but that traffic is never seen by our shapers, so to account for that we need to set a fraction that allows for that (more or less) periodic traffic. I guess one can reach a point of "goog enough" even when ignoring such eventualities, especially if having to convince through-put hot-rodders. Always interesting to hear experience from the real world, thanks!

Regards
	Sebastian

>  
> 
> 
> -Erik 
> 



More information about the Rpm mailing list