From: Muhammad Ahmed (faisal3541@hotmail.com)
Date: Tue Jun 24 2008 - 23:44:24 ART
Hello guys,
Sorry for the OT. I need some advise on a VoIP quality issue over PPP
Multilink bundle. If a question sounds stupid, its probably because it is
stupid. I do not know much about QoS and practically do not know anything
about VoIP.
I have configured the following on the PPP Multilink bundle and I just want
the to ensure there is no other configuration required to optimize VoIP
packets over the bundle.
I am using 3 T1's to form the bundle and PPP Multilink Multiclass is
configured on the serial interfaces to sequence non-fragmented VoIP packets as
well. One way delay variation between the fastest and the slowest T1 is ~8ms,
which quite frankly I do not know if 8ms is within the jitter tolerance level
for VoIP. End-to-end, one-way, latency between the phones is about 12ms
through the fastest T1 and 22ms through the slowest T1. Access-list 105 is
matching the IP addresses of the VoIP phones which have static IP addresses.
It matches any packet destined to VoIP phones at the remote site. I am hoping
matching VoIP phone IP addresses would match any and all VoIP packets that
should be treated through the QoS policy. Please confirm if this would not
match all packets. I know I should probably tag VoIP packets with DSCP or
Precedence at the VoIP phone boundary switch but I do not manage these
switches and I cannot convince the switch administrator to implement the
correct configuration fast enough. fragment delay of 1 ms results in a
fragment size of 184 bytes which I am assuming would be large enough to frame
the largest VoIP packet, please confirm. Multilink bundle is congested with
data traffic. I do know there are packet drops, for both PPP multiclass Class
0 and 1, which are probably causing the entire VoIP quality issue but since
this is my first deployment of QoS I am not confident and do not feel like I
have the command on it to definitively say that my configuration is correct
for QoS so I can start looking at resolving the drop packet issue with
whichever T1 is at fault. The total VoIP payload through the bundle is
~800kbytes during peak usage. I configured 1.544Mbps for no apparent reason
other than to avoid built in policing on the priority policy.
Best Regards and thanks for reading my misery. :)
Muhammad
BTW, the QoS policy has been removed for testing the T1's for faults so I
would not be able to provide any "show" command outputs.
class-map match-any GOLD match access-group 105
policy-map MEX_Priority class GOLD priority 1544 class class-default
fair-queue random-detect shape average percent 93 10 ms
interface Multilink1 ip address a.b.c.d 255.255.0.0 ip flow ingress ip flow
egress load-interval 30 ppp multilink ppp multilink interleave ppp multilink
group 1 ppp multilink fragment delay 1 ppp timeout multilink lost-fragment 0
200 service-policy output MEX_Priority
This archive was generated by hypermail 2.1.4 : Tue Jul 01 2008 - 06:23:23 ART