QoS

[Bob Greer]

Hi there,

Thanks for reading.

I'm implementing a QoS policy beginning with a pair of sites.  I can abide a blip but not a full-on outage or reboot.  I'm wondering about the impact of the service-policy on an interface.

Here's what i'm doing:

class-map match-any A-traffic
 match  dscp ef
class-map match-any B-traffic
 match  dscp af43
 match  dscp af31
 match  dscp cs3
class-map match-any mark_voip
 match protocol h323
class-map match-any mark_Business
 match access-group name mark_Business

policy-map Quality
 class A-traffic
    priority percent 50
 class B-traffic
    bandwidth percent 30
 class class-default
    fair-queue 256
    queue-limit 256 packets
policy-map marking
 class mark_voip
  set dscp ef
 class mark_Business
  set dscp af43

Finally, there's an ACL for mark_Business to ID that traffic.  I've already entered those commands and am only waiting to enter the service-policy command.

Thanks again for reading!

Bob

[Philip D'Ath]

Assuming the circuits are not overly congested you wont have any issue applying a service policy.

[Joseph W. Doherty · Accepted solution]

I've seldom seen installation or removal of a service-policy, itself, cause much if any service disruption.

However, the impact of adding or removing QoS can be profound, although as Phillip notes, generally not unless the link is congested (or highly utilized).

BTW, to increase the effectiveness of a service-policy, I've often found you need to decrease the interface FIFO queue size.

[Bob Greer]

Hi Philip & Joseph,

Can you point me to a document or high level guidelines?  The link is saturated so I want to double check my configuration before enabling QoS at the inside inf of my edge router. 

Thanks,

Bob

[Joseph W. Doherty]

Personally, I don't recall coming  across such documentation.

Personal experience also includes having worked on a project where we added QoS to about 5,000 enterprise class devices (mostly various L2 and L3 switches).  Again, I don't recall seeing any application issues in application to my set of devices (about 500) nor any other team members mentioning it.  We actually applied changes during business hours, although, for best practice, would recommend you make such changes during low usage or maintenance hours.

Also, before doing this project, we queried our on-site Cisco SE, who I presumed also may have queried the "mothership" - SE's answer was pretty much as already noted by Philip and myself.