|
Joined: Oct 2007
Posts: 62
Member
|
Member
Joined: Oct 2007
Posts: 62 |
Have 3 locations all with the ability to use VoIP via either UGW or PUGW. We seem to lose voice quality around 192k. Using G711 for all phones. Using Public and MPLS for phones, problem is on either network. PBX is at ver 15, all phones are ver 3.25. We do not use QoS on the public as it is Public, there is QoS on the MPLS with it being tagged and handled by the carrier. There is at least always 4mb of additional bandwidth left over before phone calls take place. Latency to/from any point never exceeds 80ms. Has anyone see anything like this or have some ideas on what can be done to improve? Can the UGW/OUGW handle more than 192k at a time?
Adrian
|
|
|
|
Joined: Jul 2006
Posts: 728
Member
|
Member
Joined: Jul 2006
Posts: 728 |
I have you made calls just between the MPLS sites and seen a drop in call quality?
rule out that portion of the network since that one is QOS driven..
Then make calls between one MPLS site and one public site. see if the call quality drops as well. I tend to believe your problem would be on the Public side that is giving you headaches.
|
|
|
|
Joined: Sep 2004
Posts: 4,220 Likes: 2
Member
|
Member
Joined: Sep 2004
Posts: 4,220 Likes: 2 |
What compression (if any) are you using?
|
|
|
|
Joined: Oct 2007
Posts: 62
Member
|
Member
Joined: Oct 2007
Posts: 62 |
Problem is the same on MPLS as well as public, magic number seems to be 192k and we see the gremlins. Currently no compression, thought this may have been problem initially. In the beginning had the call center running off of the Coral III fought tooth and nail with the carriers blaming them for these problems, ultimately changed out public carriers, then moved call center to ipx 500 at another office with new cards for that site and problem is persistent. the only 2 things that are the same is tadiran and our programming but yet the problem follows. we take probably 200,000 calls a month in the call center so it isn't a large call volume center (most business these days is web based you know) looking at the time slots, the system is barely awake yet got all kinds of TS free. over subscribed on dtmf receivers just in case of any problems. i would say at any one time we have probably 20 IP phones in use at the same time and are having these problems. using the IPG cards basically for inter office dialing we don't see these problems so the problem is unique to the UGW/PUGW. Was on the phone with joke support yesterday they of course blamed the network, we pulled the cards out of the system and still saw bandwidth utilization in the PI with no UGW/PUGW cards in the system. To me this immediately rules out network when there are no cards in the pbx and you see 76k and 96k being used but yet no cards in the system. then they wanted me to run wireshark for when the problem started, told them you can't do that you know how much data happens in 1 minute never mind that we would have to run this 8 hours and anyone that has used wireshark knows you better have one huge server to run for that amount of time or it'll die, so at that point lost my faith in tech support. If the system cannot handle it fine, if we need to change to high compression fine, if we need to reboot the pbx once a week for this to work fine, not looking to blame anyone on this just looking for resolution promise. at this point will be setting up 3 sniffers one at each office to be able to activate when we think there is a problem so we can grab that data. has anyone else had this problem? have a hard time spending another 5k for another PUGW to support 5-10 ip phones but if that is the case or what is recommended then it is what it is won't be very pleased but ultimately we just need the system to work. can you think of anything else to maybe try for troubleshooting?
Adrian
|
|
|
|
Joined: Sep 2004
Posts: 4,220 Likes: 2
Member
|
Member
Joined: Sep 2004
Posts: 4,220 Likes: 2 |
Ok, something isn't making sense here. Timeslots aren't even in the equation unless a media gateway is being utilized. Are these pure VOIP calls or is there a trunk involved? Are you using a sentinel...this could be key BTW. Also see if you are using echo cancellation on your cards.
|
|
|
|
Joined: Oct 2007
Posts: 62
Member
|
Member
Joined: Oct 2007
Posts: 62 |
PRI's are heavily used so for that there is a MGW, No Sentinel involved, Will check on echo cancellation.
Adrian
|
|
|
|
Joined: Sep 2004
Posts: 4,220 Likes: 2
Member
|
Member
Joined: Sep 2004
Posts: 4,220 Likes: 2 |
What size MGW are you using AND was your system initialized for the proper amount that you have?
|
|
|
|
Joined: Oct 2007
Posts: 62
Member
|
Member
Joined: Oct 2007
Posts: 62 |
Echo cancelation is currently set for 0 – 32 msec for both cards
yes - sized for 386 IP Keysets and 20 SIP Terminals, we are using the original IP phones not the SIP ones
Adrian
|
|
|
|
Joined: Jul 2006
Posts: 728
Member
|
Member
Joined: Jul 2006
Posts: 728 |
Adrian,
Do you have the same issue on IP phone calls in the same office?
|
|
|
|
Joined: Sep 2004
Posts: 77
Member
|
Member
Joined: Sep 2004
Posts: 77 |
I will defer to Coral and Tex as better authorities. But in my experience you need to go to G729 for better usage of your bandwidth. As I recall G711 uses 64k per link while g729 uses 16k. The question I have is what is your bandwidth limit set for on the zone you're using.
|
|
|
Forums84
Topics94,525
Posts640,011
Members49,852
|
Most Online5,661 May 23rd, 2018
|
|
1 members (Toner),
203
guests, and
43
robots. |
Key:
Admin,
Global Mod,
Mod
|
|
|
|