Yes, Skype typically tags their RTP with DSCP 46, IIRC.
The other half of the problem is DSCP originating on your transit. We
get no DSCP marked packets via Windstream. But from GTT, depending on
which of their peers the traffic is coming from, we do see DSCP. Apple
updates were coming in marked and overloading the HP channel on a couple
customers and it totally destroyed their VoIP. So check things out and
zero out DSCP/TOS coming from your upstreams if needed. Easy with a
MikroTik mangle rule. Had to do that and then pop on new tags for our
business voice which is hosted downtown Chi. This was a few years ago,
but I haven't bothered to look if anything has changed. It's working so
I ain't touchin' it.
Tim, have you worked with Cambium on this at all? It really sounds like
the max burst rate could be conflicting with HP in the scheduler.
On 4/19/2016 2:44 PM, Brian Sullivan wrote:
"make sure inbound ethernet traffic to the SM is tagged correctly so
that voip, and only voip, ends up in the high priority channel."
What other applications are tagging packets with HP? Voice chat on
gaming systems?
On 4/19/2016 1:52 PM, Timothy Alexander wrote:
We use 100 kbits per call session of G711U. We took a while to get a
good working config on junos to make sure only voip traffic comes in
to the SM tagged for high priority. The SM only goes by the Priority
Precedence selection to check either 802.1p or DSCP to determine
transmit channel and you need to make sure inbound ethernet traffic
to the SM is tagged correctly so that voip, and only voip, ends up in
the high priority channel.
We also ran into issues with burst allocation and burst rates. In
order for the high priority channel to work correctly on 450 platform
we found you must set the Sustained Uplink/Downlink Data Rate and the
Uplink/Downlink Burst Allocation to be the same (customer regular
bandwidth + high priority bandwidth) and the Max Burst
Uplink/Downlink Rates to double the Total Sustained Uplink/Downlink.
We confirmed in extensive testing that if we simply enable a high
priority channel on a normal customer, when the user maxes his upload
or download we see packet loss on the high priority channel even when
the upload/download is all low priority traffic and the high priority
channel is not being 100% used.
Why is this? We're not sure but our network engineers did pretty
extensive testing and we consistently saw packet loss when configured
with a high priority channel and a "normal" customer deployment of
Sustained Down 4 Mbit / Sustained Up 1 Mbit, Downlink Burst
Allocation 120000 and max burst downlink of 10000.
Timothy Alexander
Amplex Internet
www.amplex.net
[email protected]
From: Brian Sullivan<[email protected]>
To:"[email protected]" <[email protected]>
Subject: [AFMUG] High Priority Uplink / Downlink
What do other operators set for the High Priority Uplink / Downlink
CIR
that works best for VOIP?
