03-02-2013 07:18 PM - edited 03-07-2019 12:01 PM
Hi,
This particular switch has all ports identically configured but only has three ip phones connected. I've been asked to find ways to speed up the existing network as was wondering about this.
Does having the following configuration degrade performance for ports that do not have phones in any way?
Does the priority queue only affected marked voice packets? And if there are no voice packets, then does that delay data packets since it must first check the markings?
Thanks.
mls qos map cos-dscp 0 8 16 26 34 46 48 56
mls qos queue-set output 1 threshold 2 400 400 100 400
mls qos queue-set output 1 buffers 15 45 20 20
mls qos
(all ports identical config)
interface FastEthernet0/1
switchport access vlan 50
switchport mode access
switchport nonegotiate
switchport voice vlan 101
priority-queue out
mls qos trust cos
no mdix auto
spanning-tree portfast
Solved! Go to Solution.
03-02-2013 09:25 PM
No, it doesn't degrade the performance since the none voice packet use default cos (0).
But if you only have voice connected to 3 ports why have all the extra configs on all the pther ports. So, on the ports that you don't have any phone connected to remove below configs
switchport voice vlan 101
priority-queue out
mls qos trust cos
Also
also priority-queue out command is an egress priority queue and usually applies to uplinks and not ports connected to hosts.
HTH
ority-queue out
03-02-2013 09:42 PM
Hi.
This switch is QoS enabled, that means that queues memory allocation is modified, If your switch is pure data flow, and has minimal or nothing voice traffic, you shouldn't be using QoS.
Queues assinged to queue-set1 are limited the following way:
Queue 1 has only 15% of the total interface resources.
Queue 2 as 45%
Queue 3 and 4 has 20% each.
Take at look at the output of show interfaces summary, if you notice ODQ (output drops) it's most likely due to bad qos implementation.
Also verify if qos is really enabled, "show mls qos" can tell you that.
Does having the following configuration degrade performance for ports that do not have phones in any way?
YES!!! Image if everything is pure data flow, most likely by default data traffic is marked with a DSCP value of 0. If so data traffic is mapped to queue 1, and guess what queue 1 only takes advantage of 15% of total interface resources.
Does the priority queue only affected marked voice packets?
It will affected only traffic associated to the priority queue, by default in fixed units priority queue is 1 threshold 1.
And if there are no voice packets, then does that delay data packets since it must first check the markings?
Priority queue is serve first until empty. markings are done during queueing and scheduling and yes it could add delay, if the buffer for non-priority queues are full before the priority queue gets served then the switch will start dropping packets, causing at the same time retransmissions, packet loss, general performance issues and then things start going wrong...
Regards.
Wilson B.
Does the priority queue only affected marked voice packets? And if there are no voice packets, then does that delay data packets since it must first check the markings?
- See more at: https://supportforums.cisco.com/message/3871458#3871458
03-02-2013 09:25 PM
No, it doesn't degrade the performance since the none voice packet use default cos (0).
But if you only have voice connected to 3 ports why have all the extra configs on all the pther ports. So, on the ports that you don't have any phone connected to remove below configs
switchport voice vlan 101
priority-queue out
mls qos trust cos
Also
also priority-queue out command is an egress priority queue and usually applies to uplinks and not ports connected to hosts.
HTH
ority-queue out
03-03-2013 11:46 AM
About it affecting performance, if the priority queue is bound to an egress trunk port, and the markings give data default cos 0, and that is only allocated 15% of the port's available buffers, it seems to potential to degrade data performance is present but only if the interface commands (especially priority-queue out) are left in place.
So removing those interface commands in the case where no phone is present makes a great deal of sense.
Thanks.
03-02-2013 09:42 PM
Hi.
This switch is QoS enabled, that means that queues memory allocation is modified, If your switch is pure data flow, and has minimal or nothing voice traffic, you shouldn't be using QoS.
Queues assinged to queue-set1 are limited the following way:
Queue 1 has only 15% of the total interface resources.
Queue 2 as 45%
Queue 3 and 4 has 20% each.
Take at look at the output of show interfaces summary, if you notice ODQ (output drops) it's most likely due to bad qos implementation.
Also verify if qos is really enabled, "show mls qos" can tell you that.
Does having the following configuration degrade performance for ports that do not have phones in any way?
YES!!! Image if everything is pure data flow, most likely by default data traffic is marked with a DSCP value of 0. If so data traffic is mapped to queue 1, and guess what queue 1 only takes advantage of 15% of total interface resources.
Does the priority queue only affected marked voice packets?
It will affected only traffic associated to the priority queue, by default in fixed units priority queue is 1 threshold 1.
And if there are no voice packets, then does that delay data packets since it must first check the markings?
Priority queue is serve first until empty. markings are done during queueing and scheduling and yes it could add delay, if the buffer for non-priority queues are full before the priority queue gets served then the switch will start dropping packets, causing at the same time retransmissions, packet loss, general performance issues and then things start going wrong...
Regards.
Wilson B.
Does the priority queue only affected marked voice packets? And if there are no voice packets, then does that delay data packets since it must first check the markings?
- See more at: https://supportforums.cisco.com/message/3871458#3871458
03-03-2013 11:48 AM
Thanks much for your reply. A lot of good information there.
If a switch only has 3 out of 48 active ports VoIP, would you enable QoS on the switch then in a LAN setting?
Can you point me to a good link that explains all of these settings and/or QoS setup in general?
03-03-2013 02:30 PM
Take at look at the output of show interfaces summary, if you notice ODQ (output drops) it's most likely due to bad qos implementation.
I just checked all the switches at this site. They all have QoS implemented and the only one that doesn't have output drops is the one that doesn't have any phones connected.
Seems like you may have something there.
03-04-2013 06:58 AM
Hello again.
Unfortunately Catalyst QoS can't be explained in general, it differs from platform to platform, from supervisor to supervisor, from linecard to linecard and even IOS makes changes to the way how it's configured, The best I can advise is to check the configuration guides at cisco.com everytime you have a question about qos.
The tittle Cisco Catalyst QoS: Quality of Service in Campus Networks explains a bit of everything; classification, marking and policing, scheduling and queueing, shaping and sharing, weighted round robin and shaped round robin) .....
Regards.
Wilson B.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide