Hi
Are doing tests with a HQOS profile for L2VPN with 4 cos levels where EF and AF4 are prioritized and limited to take max 50%. The remaining bandwidth are to be shared between AF3 and AF1 in a 9 to 1 relation. For this "bandwidth remaining ratio" is used. When testing a hich load scenario with all 4 cos levels loaded and running several TCP flows in AF1 to push slightly over the top there is queue build up in AF3 and AF4 even if the load in these classes are within limit. The result is high delay variation in AF3 and even some in AF4. Have tried different different ratio settings like 9/1, 90/10, 27/3 and can se different results. So my question is what entity the "remaining bandwidth ratio" is? Is it amount of queue particles (256 bytes?) served per scheduler round? And if so is there a best practize, like not setting a lower value than max mtu size or so?
Appreciate any help I can get.
Best regards,
Martin
Here is a sample policy:
policy-map QOS_CUST_L2_OUT_10M_M
class QOS_CUST_EF_QOSGROUP
set cos 5
police rate 3000 kbps
!
priority level 1
!
class QOS_CUST_AF4_QOSGROUP
set cos 4
police rate 3000 kbps
!
priority level 2
!
class QOS_CUST_AF3_QOSGROUP
set cos 3
bandwidth remaining ratio 9
queue-limit 188 kbytes
!
class class-default
set cos 1
bandwidth remaining ratio 1
queue-limit 188 kbytes
!
end-policy-map
!
policy-map QOS_CUST_SHAPE_L2_OUT_10M_M
class class-default
service-policy QOS_CUST_L2_OUT_10M_M
shape average 10000 kbps
!
end-policy-map
!
