IP Network Regions & QOS - Question...

[phoneguy01]

Let's say you've got an OC3 circuit that goes into your WAN cloud and that single circuit feeds several sites - those sites are part of your phone system. Now, let's say on the OC3 at the main site you have a total of 1MB of Gold CAR...meaning Verizon will guarantee that 1MB of the traffic we mark GOLD will have the highest priority and not be dropped for congestion. So the issue is we have several network regions setup and each of course needs to have a bandwidth limit set. I'm thinking we can't set 7 network regions each to 1MB because if we did, then would the system assume that we're telling it that it is allowed to send a total of 7MB of calls to those sites??? Know what I mean? Those 7 sites use the same WAN circuit and that circuit only has 1MB of Gold CAR for QOS. How do we arrange this so that we can make full use of the 1MB of QOS'd traffic but not go over that limit by telling the system each network region has it's own 1MB of bandwidth? Maybe I'm not understanding this correctly. Please advise if you've done this before....same scenario. What did you do? I would assume this would be a pretty common scenario.

 

[wildcard100001]

I'd say you need to increase QOS/Gold bandwidth limit.

Run the command: stat ip-n 1 (1 being your hub network region), run it at your peak times of the day, this will give you an idea of how much bandwidth you are actually using on a call by call basis to each region. Also allow at least 150k per IPSI on top of your call QOS limits. You'll want to always have enough QOS bandwidth for IPSI traffic.
Run: list measurements ipserver-interface summary today-peak to see peak IPSI bandwidth throughput.

Wildcard
CM 5.1.1

 

[phoneguy01]

I have the ip network regions status report emailed to me at the end of every day so I'm aware of how many times IGAR was used, BW exceeded, etc. We simply need to know if we can allocate the 1MB of Gold CAR/priority bandwidth in a smart way so the sites can share that bandwidth coming in and out of the main site. Otherwise we'd be paying through the nose for lots of Gold CAR we don't need.

 

[inerguard]

What you need to remember is that QOS only becomes effective when you have a congestion scenario.

You should have all 7 sites sending info and if the traffic is priority mark it as gold.

You wont always have 1mb from each site, so you wont need 7mb in total. Say you have the following at any one particular time : -

Site 1 128k
Site 2 64k
Site 3 256k
Site 4 32k
Site 5 64k
Site 6 256k
Site 7 128k

Total is 928k. All of this traffic would not need prioritising as there isn't congestion.

If you had 2MB of Data from Site 1 as well as 128K of voice traffic for instance, then you gold mark the 128k of voice traffic from site 1 etc. All of the gold traffic is added up and prioritised only when the overall bandwidth requirements create a bottleneck and hence only then does QOS become a requirement to prioritise some poackets over others.




[Started on Version 3 software 15 years a go]

 

[phoneguy01]

I understand what you're saying about QOS not having any affect unless we have congestion but we DO mark all the traffic from the Medpro cards or G700 gateways as Gold. So the issue we have is, what happens if we say each network region (from/to the main network region where I am) is allowed to use up to 1 MB of bandwidth? I assume that the system would think we are telling it that it can use a total of 7MB of bandwidth at any one time...when we really only have 1MB of bandwidth to share among all site sites at this main site for this single circuit. Know what I mean? The real question is - is there a way to say each region can use up to 1MB of bandwidth as long as it's available...I.E., not being used by the other network regions connected/talking to the network region at the main site here? Hopefully I'm explaining this correctly

 

[phoneguy01]

Found the answer....setup a new intervening network region that all regions with the certain bandwidth limit must pass through before getting to the destination network region. That way, you can control the bandwidth since they all must pass through the same region first. You're just using that one new region to control the bandwidth.

 

[wildcard100001]

You'd still be limiting your traffic... Are you going to deny say region 3 because region 2 is having a calling storm? I'd pony up the $$$... size it to your theoretical max possible traffic. Though it depends on what is "good enough" to your company and perhaps what type of call traffic it is. I have call center traffic going all over the place to many locations. I would not do what you are doing in my situation. I also have centralize voicemail etc..

inerguard - I would not agree with your congestion comment. VOIP traffic should always have QOS even if you have 10g pipe. I'm not saying it won't work, but quality will eventually suffer.

Wildcard