[Cake] Recomended HW to run cake and fq_codel?

[Sebastian Moeller]

Hi Erik,

> On May 3, 2017, at 09:27, <erik.taraldsen at telenor.com> <erik.taraldsen at telenor.com> wrote:
> 
>> Fra: Sebastian Moeller <moeller0 at gmx.de>
>> 
>>       Question: as an ISP what is your rationale to implement a shaper at the BRAS? Simply the fact that 
>> DSLAMs/MSANs are not capable to do it, or do you also need this to make sure there is always room for > your own VoIP packets?
> 
> At least the DSLAMs we use are basically switches.  Extremly limited QoS/shaping.  For a time we actually did not shape at the BRAS level and let the DSLAM's just drop whatever it could not push throug.  That was not a success.  It more or less behaved like a strictly policed access, which is not something you want.  So we went back to shaping at the BRAS.

	Ah, thanks for the insight, I had always assumed the rationale to be less obvious (like making VoIP more robust, keeping (D)DOS traffic out of the aggregation network; I had never assumed that dslams might simply be bad at traffic regulation ;) )

> 
> 
> 
>> 0) get rid of all non-essential encapsulations, use DHCP option 82 instead of PPPoE, rethink the need 
>> for VLAN tags,
> 
> Done, with the exeption of some legacy business usages.

	This is quite enlightened, great!

> 
> 
>> 1) make sure to properly account for all the quirks of ATM’s AAL5 encapsulation (see cake’s 
>> atm keyword or “man tc-stab”)
> 
> Noted.

	All I ever learned about this topic should be summarized in https://github.com/moeller0/ATM_overhead_detector/wiki .

> 
>> 2) preferably hoist your ADSL customers into the present and get your device manufacturers 
>> to implement PTM for adsl modems making 1) above much less involved ;)
> 
> To much legacy, so  more likely to migrate to VDSL across the board.

	Which effectively is the same for customers (except for unlucky ones at the end of very long lines, I believe); ATM deserves to retire ;)


> 
> Regarding BQL, we need the chipset vendors to do this.  In particular Broadcom.  We do try and influence them to do this, but we simply are a to small to get traction.

	Well, compared to most on this list you have a huge impact on the chipset vendors, so I hope for the best. Is there anybode besides Broadcom and Intel actually producing VDSL chipsets or is that the set of vendors that need to be convinced?

Best Regards
	Sebastian

> 
> 
> -Erik