Discards/QOS Drops on WAN facing interfaces

Ableton34 · ‎05-18-2017

Hi all,

Wondering if you can help. Can you explain if Ethernet WAN links should cause discards/QOS drops in a normally functioning link/circuit when the QOS is correctly configured at each end of the CE and PE router into a cloud with the correct shaping. Should we even tolerate any drops in the QOS or is this considered normal?

An example from one of my links:

GigabitEthernet0/0/0

Service-policy output: WAN-QUEING-PARENT

Class-map: SHAPE-ALL (match-all)
1464646054 packets, 417358598058 bytes
30 second offered rate 790000 bps, drop rate 0000 bps
Match: any
Queueing
queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/119493/0
(pkts output/bytes output) 1463413294/415499768625
shape (average) cir 10000000, bc 40000, be 40000
target shape rate 10000000

Service-policy : WAN-QUEING-CHILD

queue stats for all priority classes:
Queueing
queue limit 512 packets
(queue depth/total drops/no-buffer drops) 0/0/0
(pkts output/bytes output) 14751887/3582918036

Class-map: DACP_PREMIUM_QUEING (match-all)
14757148 packets, 3590886074 bytes
30 second offered rate 3000 bps, drop rate 0000 bps
Match: dscp ef (46)
Priority: 1000 kbps, burst bytes 25000, b/w exceed drops: 0

Class-map: DACP_ENHANCED2_QUEING (match-all)
167659987 packets, 11081680031 bytes
30 second offered rate 9000 bps, drop rate 0000 bps
Match: dscp af31 (26)
Queueing
queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/2801/0
(pkts output/bytes output) 167652145/11073486512
bandwidth 1500 kbps

Class-map: class-default (match-any)
1282229232 packets, 402686105885 bytes
30 second offered rate 775000 bps, drop rate 0000 bps
Match: any

queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/116692/0
(pkts output/bytes output) 1281009262/400843364077

Class-map: class-default (match-any)
0 packets, 0 bytes
30 second offered rate 0000 bps, drop rate 0000 bps
Match: any

queue limit 416 packets
(queue depth/total drops/no-buffer drops) 0/0/0
(pkts output/bytes output) 0/0

configured as:

class-map match-all SHAPE-ALL
match any
class-map match-all DACP_ENHANCED2_QUEING
match dscp af31
class-map match-all DACP_PREMIUM_QUEING
match dscp ef
!
policy-map WAN-QUEING-CHILD
class DACP_PREMIUM_QUEING
priority 1000
class DACP_ENHANCED2_QUEING
bandwidth 1500
policy-map WAN-QUEING-PARENT
class SHAPE-ALL
shape average 10000000
service-policy WAN-QUEING-CHILD

int gi0/0/0

service-policy output WAN-QUEING-PARENT

This is a 10mb circuit where the traffic volumes do not appear to reach that level of usage at any time.

thanks in advance

Joseph W. Doherty · ‎05-18-2017

Often QoS on high bandwidth links needs adjustment for their queue depths; defaults often are too small for the BDP. Even though your usage level doesn't appear high, transient bursts can overflow too small queues.

If your WAN link is 10 Mbps, but your physical hand-off is higher, you need to shape, but most Cisco shapers don't appear to account for L2 overhead. Often if you shape about 15% slower, you're in the ballpark. Or, if possible, you might run your physical interface at 10 Mbps, avoiding the need for a shaper.