Cable Company Hotspots

Rob Seastrom rs-lists at
Mon Nov 23 13:55:37 UTC 2020

On Nov 22, 2020, at 12:42, Lady Benjamin PD Cannon <ben at> wrote:
> Rod, that’s exactly how they are delivering it. Unclear wether it’s over a separately provisioned bandwidth channel, or wether it shares the aggregate capacity of the HFC.

It shares the aggregate bandwidth of the HFC but not your contracted bandwidth.  Itmight be possible, but its extremely unlikely, to dedicate downstream or particularly upstream DOCSIS channels for this, and if you’re running docsis 3.1 “channel” takes on a rather different shade of meaning anyway.

This is done with “service flows” which are part of the docsis spec.  They’re more like CAR with an ACL than DSCP.  Your cable modem already has at least four service flows defined in its profile:  one each for upstream and downstream, cablemodem management and contracted-bandwidth commodity internet.   If there is a built in phone jack (NANOG would call this an ATA, but the cablelabs term for it is an MTA or eMTA) then add a couple of more flows to it for the voip.  There could be still more; uses are up to your imagination.

I haven’t seen better than 10-20m service flows for guest wifi...

Shared vs dedicated wifi radio for guest would be dependent on the CPE.  I believe they are mostly shared, but my information is dated at this point and radios have gotten stupid cheap in the meantime.

Likewise, backhaul technology is implementation dependent; L2TP is what I’ve generally seen, not GRE, but again that info is five years out of date at this point.

So in short, assuming minimal interference and good wifi config (which may be a lot to ask in some environments) someone running speedtest on the guest wifi should have almost no effect on your contracted network performance, modulo any timing effects of the docsis channel transmission time slot allocator.


Sent from my iPad

More information about the NANOG mailing list