i am looking to configure QoS on a customers 4331 ISR. Basic 'router on a stick' setup with a basic Ethernet port-channel configured on three 1G interfaces. Port-channel contains about 10 dot1q sub interfaces.
The policy I'm looking to configure is below:
policy-map pm-queue-out class cm-class-network bandwidth remaining percent 2 class cm-class-multimedia5 priority percent 33 class cm-class-multimedia4 bandwidth remaining percent 40 class cm-class-interactive3 bandwidth remaining percent 30 fair-queue class cm-class-interactive2 bandwidth remaining percent 15 fair-queue class cm-class-data1 bandwidth remaining percent 8 fair-queue class class-default bandwidth remaining percent 5 fair-queue set dscp default
I've looked through Cisco documentation which all points to configuring aggregated QoS:
The router does not currently have port-channel load-balancing vlan configured which is per-requisite for this.
Is anyone able to point me in the right direction? Do I just apply the inbound marking and outbound policy-maps to the sub-interfaces? If I want to shape traffic on a particular sub-interface is it as simple as applying a child shaping policy to that particular interface?
Thanks Balaji, I don't know if that is specific port-channel configuration guidance and I think it may have changed a bit since IOX-XE?
I understand I can apply the shaping policy to the Po1.100 sub-interface but I wanted to understand how that would work in-conjunction with the rest of the member interfaces which only have the child policy configured - if that makes sense.
Sorry, I'm a bit confused. Are you just trying to manage the whole Etherchannel as a single interface, manage each subinterface, individually, with some allocated proportion of bandwidth per subinterface or trying to manage individual Etherchannel links?
BTW, managing per Etherchannel link, would be rather unusual.
Managing subinterfaces can be difficult for making effective QoS unless you set aside a fixed proportion of bandwidth for the exclusive, and also always available, use of each subinterface.
For example, you have:
class cm-class-multimedia5 priority percent 33
and you mention having about 10 subinterfaces. So, is the 33% for all 10 subinterfaces using this kind of traffic, or 33% of each subinterface's allocated bandwidth?
Also BTW, since you're using FQ in many of your classes, you might have twice (or more) classes than you really need.
Often I suggest starting with a CBWFQ model based on:
class real-time !for traffic like VoIP (bearer) priority percent 35
class foreground bandwidth remaining percent 81 !light bandwidth usage traffic that really, really needs priorization fair-queue
class background bandwidth remaining percent 1 !high bandwidth, low prioity traffic, that is "happy" with left over bandwidth fair-queue
class class-default !most non-real-time traffic goes here - it's FQ which, often, negates much need for other classes than RT bandwidth remaining percent 9 fair-queue