05-17-2013 12:50 PM - edited 03-04-2019 07:56 PM
Hello,
I have a 2921 on Ethernet MPLS circuit. Problem is Voice has jitter at 60ms and no dropped packets from source to destination. How to reduce the jitter? Is the policer correct using Cisco recomended Nb = CAR x (1/8) x 1.5?
The PE is honoring the DSCP marked packets.
CE router 2921 QOS:
policy-map IFCQOS
class EF
priority 2048
set dscp ef
police 2048000 384000 768000
class AF4
bandwidth remaining percent 70
set dscp af41
class ProtocolTraffic
bandwidth remaining percent 5
class class-default
fair-queue
set dscp 2
policy-map shape-etm-IFCorp
class class-default
shape average 9600000
service-policy IFCQOS
Class-map: EF (match-any)
1927181 packets, 219660148 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: access-group name EF
1927181 packets, 219660148 bytes
5 minute rate 0 bps
Priority: 2048 kbps, burst bytes 51200, b/w exceed drops: 0
QoS Set
dscp ef
Packets marked 1927181
police:
cir 2048000 bps, bc 384000 bytes, be 768000 bytes
conformed 1927181 packets, 219678100 bytes; actions:
transmit
exceeded 0 packets, 0 bytes; actions:
drop
violated 0 packets, 0 bytes; actions:
drop
conformed 0 bps, exceed 0 bps, violate 0 bps
policy-map IFCQOS
class EF
priority 2048
set dscp ef
police 2048000 384000 768000
class AF4
bandwidth remaining percent 70
set dscp af41
class ProtocolTraffic
bandwidth remaining percent 5
class class-default
fair-queue
set dscp 2
policy-map shape-etm-IFCorp
class class-default
shape average 9600000
service-policy IFCQOS
Class-map: EF (match-any)
1927181 packets, 219660148 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: access-group name EF
1927181 packets, 219660148 bytes
5 minute rate 0 bps
Priority: 2048 kbps, burst bytes 51200, b/w exceed drops: 0
QoS Set
dscp ef
Packets marked 1927181
police:
cir 2048000 bps, bc 384000 bytes, be 768000 bytes
conformed 1927181 packets, 219678100 bytes; actions:
transmit
exceeded 0 packets, 0 bytes; actions:
drop
violated 0 packets, 0 bytes; actions:
drop
conformed 0 bps, exceed 0 bps, violate 0 bps
Solved! Go to Solution.
05-19-2013 02:21 AM
Hello,
You don't show what your EF class is classified against - but it would be best to
separate voice/video traffic in different classes so if either exceeds its limit that class will only be affected not both.
As for setting tc, Cisco recommendations for sensitive applications is 0.010
TC= 0.125 default
TC= 0.025 for lower 320kps
TC= 0.010 for sensitive applications
So to shape at 4mb this would has a bc of 5120
shape average 4096000 5120
res
Paul
Please don't forget to rate any posts that have been helpful.
Thanks.
05-19-2013 02:37 AM
Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Okay, I understand about 10 Mbps logical limit being a L2 limit, and about most low-end shapers (I believe) don't account for L2 overhead. Again, if possible, if you're able to set the physical interface to 10 Mbps, you don't need to shape. Otherwise, yes you shape for 10 Mbps less your expected L2 overhead. (For example I think your 9.6 Mbps is too large for VoIP packets.) Do keep in mind, L2 overhead is variable based on packet size. This mean either you need to shape for worst case or your average case.
I think your Bc for Tc is formula is correct (not at a device to check), although you don't want 400ms (.4) but 4ms (.004) (or perhaps 10ms [.01)]. Somewhere in show the shaper stats, often Tc being used is shown. You don't need to set Be as average shaping doesn't use it and I believe it defaults to whatever you set Bc to.
05-17-2013 05:08 PM
Hello
Personally I wouldn't police andLLQ for the same class
Try removing the policer and see if the jitter is resolved
Res
Paul
Sent from Cisco Technical Support iPad App
05-18-2013 04:29 AM
Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Ok, the PE honors your DSCP marked packets - which means exactly what across their MPLS cloud?
Problem might not be on your side at all. However, some things to consider . . .
If your CAR is 10 Mbps, QoS works best against physical interfaces. I.e. instead of using a shaper, run the interface (if possible) at 10 Mbps.
Shapers, especially dealing with VoIP, usually need to have their Tc at 10ms or below. I would image a 2921 is running 15x, and I believe the later IOSs will default to a Tc of 4ms, but you should check your shaper's stats.
You may also want to reduce the TX-ring-limit for your physical interface (although newer IOSs are supposed to better set it).
Why do you have an explict policer in your LLQ class?
Regarding your question about your policer's parameter settings, "it depends". Policers (and shapers) are trying to simulate the amount of traffic that a physical interface would pass during some time interval. Behavior, depends much on time interval involved and the also the amount of buffering the simulated physical interface would have. Getting semi-accurate simulation behavior can be rather complex, so CAR or CIR x 1.5 is just a very rough rule-of-thumb as not allowing for some policer bursting tends to be very detrimental to bursty traffic.
05-18-2013 07:08 AM
Thx for the reply. I removed the policer and jitter is the same. The shaper is in there to compensate for Ethernet Overhead. I used the policer to compensate for excessive queuing but now realize Bc should have been in the shaper.
Will making an adjustment to the policy help with Jitter? I keep reading to set the Tc to .5 for voice but not shure how to do that. I think it is adding Bc to the shaper. Something like: shape average 9600000 600000?.
policy-map IFCQOS
class EF
priority 2048
set dscp ef
class AF4
bandwidth remaining percent 70
set dscp af41
class ProtocolTraffic
bandwidth remaining percent 5
class class-default
fair-queue
set dscp 2
policy-map shape-etm-IFCorp
class class-default
shape average 9600000
service-policy IFCQOS policy-map IFCQOS
class EF
priority 2048
set dscp ef
class AF4
bandwidth remaining percent 70
set dscp af41
class ProtocolTraffic
bandwidth remaining percent 5
class class-default
fair-queue
set dscp 2
policy-map shape-etm-IFCorp
class class-default
shape average 9600000
service-policy IFCQOS
wbg-ifckiev-2728891#sh policy-map int
GigabitEthernet0/0
Service-policy output: shape-etm-IFCorp
Class-map: class-default (match-any)
3874897 packets, 2206585782 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: any
Queueing
queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/26098/0
(pkts output/bytes output) 3848799/2169656564
shape (average) cir 9600000, bc 38400, be 38400
target shape rate 9600000
Service-policy : IFCQOS
queue stats for all priority classes:
queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/0/0
(pkts output/bytes output) 1931741/220124980
Class-map: EF (match-any)
1931741 packets, 220043188 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: access-group name EF
1931741 packets, 220043188 bytes
5 minute rate 0 bps
Priority: 2048 kbps, burst bytes 51200, b/w exceed drops: 0
QoS Set
dscp ef
Packets marked 1931741
Class-map: AF4 (match-any)
531446 packets, 720562938 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: access-group name AF4
531446 packets, 720562938 bytes
5 minute rate 0 bps
Queueing
queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/0/0
(pkts output/bytes output) 531446/720644836
bandwidth remaining 70% (5286 kbps)
QoS Set
dscp af41
Packets marked 531446
Class-map: ProtocolTraffic (match-all)
8382 packets, 1537588 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: access-group name BGP
Queueing
queue limit 64 packets
(queue depth/total drops/no-buffer drops) 0/0/0
(pkts output/bytes output) 8382/665530
bandwidth remaining 5% (377 kbps)
Class-map: class-default (match-any)
1403328 packets, 1264442068 bytes
5 minute offered rate 0 bps, drop rate 0 bps
Match: any
Queueing
queue limit 64 packets
(queue depth/total drops/no-buffer drops/flowdrops) 0/26098/0/26098
(pkts output/bytes output) 1377230/1228221218
Fair-queue: per-flow queue limit 16
QoS Set
dscp 2
Packets marked 1392851
05-18-2013 05:35 PM
Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Why do you need to compensate for Ethernet overhead? Is your physical port speed faster than 10 Mbps but you have a 10 Mbps cap?
Yes, you set Tc by setting Bc. If you don't set it, it will default. Newer IOSs, I believe, default to 4ms. Half a sec is much too large.
05-18-2013 09:20 PM
Yes exactly. Circuit is sold at for example 10M. Customer Edge router will need to account for the overhead so not to overutilize upstream ports such at a signaling converter (Overture, Anda) and the provider edge port.
Think I have it now or maybe my math is off a bit. So to get Tc .4 the calcultation is: Nb = CAR x (1/8) x .4. or 4096000 x 1/8 x .4 = 204,800, shape average 9600000 480000 480000 ?
Thanks for you posting. Very helpful.
05-19-2013 02:21 AM
Hello,
You don't show what your EF class is classified against - but it would be best to
separate voice/video traffic in different classes so if either exceeds its limit that class will only be affected not both.
As for setting tc, Cisco recommendations for sensitive applications is 0.010
TC= 0.125 default
TC= 0.025 for lower 320kps
TC= 0.010 for sensitive applications
So to shape at 4mb this would has a bc of 5120
shape average 4096000 5120
res
Paul
Please don't forget to rate any posts that have been helpful.
Thanks.
05-19-2013 02:37 AM
Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Okay, I understand about 10 Mbps logical limit being a L2 limit, and about most low-end shapers (I believe) don't account for L2 overhead. Again, if possible, if you're able to set the physical interface to 10 Mbps, you don't need to shape. Otherwise, yes you shape for 10 Mbps less your expected L2 overhead. (For example I think your 9.6 Mbps is too large for VoIP packets.) Do keep in mind, L2 overhead is variable based on packet size. This mean either you need to shape for worst case or your average case.
I think your Bc for Tc is formula is correct (not at a device to check), although you don't want 400ms (.4) but 4ms (.004) (or perhaps 10ms [.01)]. Somewhere in show the shaper stats, often Tc being used is shown. You don't need to set Be as average shaping doesn't use it and I believe it defaults to whatever you set Bc to.
05-20-2013 06:54 AM
Thanks everyone for posting. Very helpful.
I believe question is answered by 2 poeple and Im good now. THANKS!
I agree 96% is to high. I'm going to lower it 85% see if that help. Also looks like when I set Tc to 0.01 the router auto adjusts to Tc 0.032ms
routername removed: (config-pmap-c)# shape average 9600000 12000 12000
Shaping Interval is 1 milliseconds. Intervals below 4 milliseconds tuned
Posting for reference:
ip access-list extended AF4
permit ip any any precedence flash
permit ip any any dscp af41
ip access-list extended BGP
permit ip any any precedence internet
permit tcp any any eq bgp
permit tcp any eq bgp any
ip access-list extended EF
permit ip any any precedence critical
permit ip any any dscp ef
!
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