HTB
HTB is an alternative to CBQ (lower CPU usage & better help)
Here is a script to optimise one end of an IAX over SDSL link:
#!/bin/sh
TCOP="add"
IPTOP="-A"
if [[ "$1" == "stop" ]; then
echo "Stopping..."
TCOP="del"
IPTOP="-D"
fi
# +---------+
# | root 1: |
# +---------+
# |
# +----------------------------+
# | class 1:1 |
# +----------------------------+
# | | |
# ~np~+----+~/np~ ~np~+----+~/np~ ~np~+----+~/np~
# ~np~|1:10|~/np~ ~np~|1:20|~/np~ ~np~|1:30|~/np~
# ~np~+----+~/np~ ~np~+----+~/np~ ~np~+----+ ~/np~
# |
# ~np~+--------+--------+~/np~
# | | |
# ~np~+-----+~/np~ ~np~+-----+~/np~ ~np~+-----+~/np~
# ~np~|1:100|~/np~ ~np~|1:101|~/np~ ~np~|1:102|~/np~
# ~np~+-----+~/np~ ~np~+-----+~/np~ ~np~+-----+~/np~
# 1:10 is the class for VOIP traffic, pfifo qdisc
# 1:20 is for bulk traffic (htb, leaves use sfq)
# 1:30 is the class that interactive and TCP SYN/ACK traffic (sfq qdisc)
# 1:20 is further split up into different kinds of bulk traffic: web, mail and
# everything else. 1:100-102 fight amongst themselves for their slice of excess
# bandwidth, and in turn 1:10,20 and 30 then fight for any excess above their
# minimum rates.
# which interface to throw all this on (DSL)
IF=eth2
# ceil is 75% of max rate (768kbps)
# rate is 65% of max rate
# we don't let it go to 100% because we don't want the DSL modem (Pairgain MegaBit Modem 300S)
# to have a ton of packets in their buffers. *we* want to do the buffering.
RATE=576
CEIL=640
#RATE=450
#CEIL=500
tc qdisc ${TCOP} dev ${IF} root handle 1: htb default 102
tc class ${TCOP} dev ${IF} parent 1: classid 1:1 htb rate ${RATE}kbit ceil ${CEIL}kbit
tc class ${TCOP} dev ${IF} parent 1:1 classid 1:10 htb rate 64kbit ceil ${RATE}kbit prio 1
tc class ${TCOP} dev ${IF} parent 1:1 classid 1:20 htb rate 64kbit ceil ${RATE}kbit prio 2
tc class ${TCOP} dev ${IF} parent 1:20 classid 1:100 htb rate ${RATE}kbit
tc class ${TCOP} dev ${IF} parent 1:20 classid 1:101 htb rate ${RATE}kbit
tc class ${TCOP} dev ${IF} parent 1:20 classid 1:102 htb rate ${RATE}kbit
tc qdisc ${TCOP} dev ${IF} parent 1:10 handle 10: pfifo
tc qdisc ${TCOP} dev ${IF} parent 1:100 handle 100: sfq perturb 10
tc qdisc ${TCOP} dev ${IF} parent 1:101 handle 101: sfq perturb 10
tc qdisc ${TCOP} dev ${IF} parent 1:102 handle 102: sfq perturb 10
tc filter ${TCOP} dev ${IF} parent 1:0 protocol ip prio 1 handle 1 fw classid 1:10
tc filter ${TCOP} dev ${IF} parent 1:0 protocol ip prio 4 handle 4 fw classid 1:100
# IAX2 prio 0.
iptables -t mangle ${IPTOP} PREROUTING -p udp -m udp --dport 4569 -j MARK --set-mark 0x1
iptables -t mangle ${IPTOP} PREROUTING -p udp -m udp --dport 4569 -j RETURN
# everything else goes into lowest priority (best effort).
iptables -t mangle ${IPTOP} PREROUTING -j MARK --set-mark 0x4
iptables -t mangle ${IPTOP} OUTPUT -j MARK --set-mark 0x4
NB The other end of this link is controlled by a Cisco – see QoS Cisco IOS
See Also
- QoS Linux with HFSC
- Asterisk QoS
- Traffic Shaping for VOIP With Linux and FWBuilder
- VoIP Spear QoS monitoring — for 24x7x365 monitoring of your Internet’s QoS. (dead link)
- Linus ADSL Bandwidth management HOWTO (outdated)
- QoS And Traffic Shaping For VoIP Users Using iproute2 And Asterisk
Go back to QoS: Quality of Service in VOIP Networks