Hi,
I have a question concerning QoS and tunnels (on a routed interface), which are overbooking the
interface-bandwidth
Assume, I have a interface with 10 Mbit/s and 15 tunnels with a bandwith of 2 Mbit/s each.
So the overall bandwidth of the tunnels is 30 Mbit/s - means the pyhsical interface is really overbooked.
Next assuming is, that each tunnel has traffic from 1 Mbit/s average (=15 Mbit/s).
So here is a congestion of the physical interface and not for tunnels itself.
The shaper for the tunnels is "shape average 2000000" (see example conf below); so the question is, which tunnel-traffic will be dropped?
Is this: the first will win - meaning in the example, the traffic from tunnel 1-10 will pass and the traffic from tunnel 11-15 will be dropped?
Next question is, if some tunnels (still 1 Mbit/s traffic on each tunnel) have
child-policies
with QoS queues - what will QoS do, or is QoS not triggered? Remeber, the tunnel itself has "capacity", but the physical interface not.
!------------------------------------
policy-map TUN-QOS-CHILD
class NET
bandwidth percent 4
class VOICE
priority percentc 10
class DATA1
bandwidth percent 40
class DATA2
bandwidth percent 20
class BE
bandwidth percent 25
!
policy-map TUNNEL-OUT
class class-default
shape average 2000000
!
policy-map TUNNEL-WITH-QOS
class class-default
shape average 2000000
service-policy TUN-QOS-CHILD
!
interface Tunnel1
bandwidth 2000
service-policy output TUNNEL-OUT
!
! same config for tunnel 2 - 10
!
!
interface Tunnel11
bandwidth 2000
service-policy output TUNNEL-WITH-QOS
!
! same config for tunnel 12 - 15
!
!------------------------------------
Can you help / explain, or know you any whitepaper which cover this?
Best regards,
Thomas
