02-15-2007 10:04 AM
Hello,
I've configured QOS for VOIP traffic on my 2610 router. Already, I've seen some encouraging improvement with what I've implemented, but on heavy and persistent upstream traffic there is jitter for the recieving end of outbound calls. My set up is like this;
- Circuit is DSL (1.5/384). There are two routers in the set up, one is the LAN router (which does router-on-a-stick for multiple VLANs) and a WAN router. They are separated via an Adtran 800 (used for training purposes. Because the Adtran emulates a full T1 provision, the bandwidth constraint would not occur here, bottleneck would still be the DSL upstream 384.)
- The LAN router is connected to the WAN router via serial interface, which is where the QOS policies are configured. Here is the policy;
class-map match-all dscp-ef
match ip dscp ef
policy-map queue-on-dscp
class dscp-ef
priority percent 75
interface Serial0/0
bandwidth 294
backup delay 5 10
backup interface Dialer0
ip address 192.168.50.2 255.255.255.0
encapsulation ppp
ip ospf cost 1
service-module t1 timeslots 1-24
service-policy output queue-on-dscp
Right now, there isn't a rule specified for default class traffic, which according this set up would be anything that isn't classified as dscp ef. What I would like to know is; what would be the best way to give VOIP traffic absolute priority in this situation? Someone has suggested reducing MTU size to make use of higher interleaving. In this case, is this something I set up only on this router? Do I need to configure the WAN router for this approach? I haven't configured this before. Can someone recommend a document?
-Shikamaru
02-15-2007 07:00 PM
You will most likely have to implement shaping on the LAN router facing the WAN router. With the config you have listed above, you are able to send a full 1.544Mbps out the link, but the upstream router only has 384K up. To ensure you prioritize the traffic you have to be congested on the interface. To force this, create a nested policy where you shape in the parent policy to something less than 384Kbps. Then, under the class-default of the parent policy, place a child policy that places the RTP traffic into the appropriate priority queue and the control into a CBWFQ. Once your outbound traffic exceeds the shaping value, the queueing mechanism will be enforced and you should see a drastic improvement on voice quality as it relates to the outbound side. This is documented in the Teleworker SRND.
http://www.cisco.com/application/pdf/en/us/guest/netsol/ns171/c649/ccmigration_09186a008074f24a.pdf
02-15-2007 10:42 PM
Jim,
Thanks for replying. It's going to take some time to sort through this technique. I haven't heard of "nested" and "parent" policies (although I may know them by different names). Can you give an distilled example of how the syntax is set up? You mention that a child policy should be placed under the parent policy for the class default. I take this to mean;
policy-map queue-on-dscp
class dscp-ef
priority percent 75
class class-default
"syntax for placing RTP traffic"
I don't understand "control into a CBWFQ". I will look at the documentation you recommended, but can you help me understand in an example what you described?
-Shikamaru
Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: