[Asterisk-Users] What is acceptablenetworklatencyforvoipconnection?

[Damon Estep]

> How does an ISP provide a Jitter SLA on a Data T1? Jitter < 5ms? How
does
> one measure that?

You can get a good feel for delay and jitter just by running a
continuous ping to a core router on your ISPs network during peak
times(or to Google for that matter) and visually monitoring the results.
A good, unsaturated link will have extremely consistent response times
with less than 5ms variations between packets.

There are several ways to measure IP performance;

http://www.cisco.com/warp/public/126/saa.html
http://www.cisco.com/en/US/tech/tk648/tk362/tech_brief0900aecd801752ec.h
tml

Not using Cisco routers? Hunt the web for a Linux equivalent (anyone
know of one?), or;

Use any program capable of sending ICMP echoes and recording the
response times.

Gather ~24 hours of data at a time, every 1 second or more (every 1
second would be 86400 data points per day).

Load the data in a spreadsheet (or create a script to calculate)

Find the absolute value of the difference between every packet and the
packet preceding it, and then average all of the difference values.

The result is your "average jitter"

Qualify your results by disclosing the parameters you use to measure.

Include the measurement technique and expected results in your SLA so
your curious customers can verify your performance.

Monitor your network continuously, IP networks are very dynamic.

With all of that being said, most ISP customer's measure with their ear
by answering the question "is my VoIP working?"


> 
> We're an ISP, we've been doing T1's for many years. I know that
customers
> can ping any equipment or servers within our network with < 10ms
response
> times, we link with 3 large Tier 1 providers, DS3 speeds for 2 of them
and
> a
> 90mbs NMLI for the third, we're not over-saturated at all, bandwidth
to
> spare, and our customers generally report 50ms average response times
out
> to
> the internet. To further regions and when going through a couple of
> networks, it can be up to 80ms. The only time I've seen over 100ms is
to
> international destinations, but the ping response times generally stay
> consistent, no dramatic spikes. That's with a standard frame relay T1.
> Point
> to Point T1's are slightly better. However I didn't create or design
nor
> do
> I maintain this company's network, I just do the VOIP thing, so I'm
quite
> curious to see their SLA, I wonder if a misconfiguration somewhere
could
> affect quality and give me a head-ache.
> 
> For a few of our heavy VOIP customers, we use 2 Point-To-Point T1's
from
> our
> location to theirs, we then combine traffic to go across both T1's, if
1
> T1
> fails, it'd all automatically go across the remaining T1. They can
ping
> our
> Asterisk server with an average of 10ms latency. We also tend to use
> private
> IP's, to avoid internet DDOS's and worms and such. We've been able to
push
> 35-40 calls across this link. Probably more if I used trunking and/or
> switched codecs from ULaw.
> 
>