[155134] in North American Network Operators' Group
Re: Rate shaping in Active E FTTx networks
daemon@ATHENA.MIT.EDU (Phil)
Fri Jul 27 08:49:59 2012
In-Reply-To: <E13755D5-359A-41D0-A539-7B402BEE3BA7@lixfeld.ca>
From: Phil <bedard.phil@gmail.com>
Date: Fri, 27 Jul 2012 08:49:05 -0400
To: Jason Lixfeld <jason@lixfeld.ca>
Cc: "nanog@nanog.org" <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On the downstream end the limiting is usually done on the subscriber aggrega=
tion equipment. Router vendors sell linecards with large amounts of queue c=
apability for this reason. This is where you would introduce some kind of Q=
oS to deal with video or voice as well. Upstream could be done the same wa=
y if they have true direct connections to the gear or be done on a CPE.=20
As far as differentiating traffic within an Internet pipe that is a slippery=
legal slope. Others have mentioned the bigger players like Procera and San=
dvine. =20
Phil
On Jul 26, 2012, at 3:45 PM, Jason Lixfeld <jason@lixfeld.ca> wrote:
> Hi all,
>=20
> I'm trying to gauge what operators are doing to handle per-subscriber Inte=
rnet access PIR bandwidth in Active E FTTx networks. =20
>=20
> I presume operators would want to limit the each subscriber to a certain P=
IR, but within that limit, do things like perform preferential treatment of i=
nteractive services like steaming video or Skype, etc., ahead of non-interac=
tive services like FTP.
>=20
> My impression is that a subscriber's physical access in these networks is e=
xponentially larger than their allocated amount of Internet access. This wo=
uld leave ample room on the physical access access for other services like V=
oice and IPTV that might run on separate VLANs than the Internet access VLAN=
. That said, I doubt there's really that much of a concern about allocating P=
IR on these other service VLANs.
>=20
> So in terms of PIR for Internet access, is there some magic box that sits b=
etween the various subscriber aggregation points and the core, which takes c=
are of shaping the subscriber's Internet access PIR, while making sure that t=
he any preferential treatment of interactive services is performed.
>=20
> Is that a lot to ask for one box? The ridiculously deep buffers required i=
n order to shape to PIR vs. police to it (because policing to a PIR is just p=
lain ugly) and the requirements to perform any sort of preferential packet t=
reatment above and beyond that seem like quite a lot to ask of one box. Am I=
wrong?
>=20
> Who might make a box like this, if it exists? And if not, what are folks u=
sing the achieve these results?
>=20
> Thanks in advance for any insights..
