[ippm] RE: [AVT] G.1050

["Alan Clark" <alan.d.clark at telchemy.com>]

To provide some background to ITU-T G.1050, for general AVT and IPPM
interest (at least I hope it is interesting).

As you know, IP impairments are strongly time varying.  This means that to
develop useful metrics you either need to sample frequently or have metrics
that can represent the time varying nature of the data (e.g. RFC3611's
burst/gap metrics and the IPPM equivalents).  Several years ago I was
exploring the use of time series models as a means of measuring and
expressing packet delay variation, and developed a simple model based on
bursts of impulses fed into a filter function. 

At around the same time, we were getting frustrated with IP network
emulators that insisted that PDV was either Gaussian or uniformly
distributed, and for our work on Voice/Video over IP performance analysis
needed IP impairment emulation/ simulation tools that were more realistic.
We started to use the time series model as a way of generating time varying
IP impairments, really for internal use, and showed this to a few standards
groups as an interesting approach.  The US TIA TR30.3 committee was just
embarking on a project to develop an IP network model for use in testing
V/M/FoIP and liked the approach - the work was later taken into ITU and
became G.1050.

We "open sourced" the software as we consider this to be of general interest
and wanted to encourage development of new IP network emulation/simulation
algorithms.

I'm not sure that AVT is the appropriate group to discuss implementation
issues in an ITU Recommendation however would certainly welcome any
discussion with anyone interested in this topic.  I think there is plenty of
scope to improve, extend, refine the approach, do further validation in
different types of network .........

Regards

Alan



-----Original Message-----
From: Steve Underwood [mailto:steveu at coppice.org] 
Sent: Monday, March 19, 2007 9:23 AM
To: avt at ietf.org
Subject: [AVT] G.1050

Hi all,

I tried implementing the IP network model in ITU G.1050, but found a few
things unclear. I tried our old friend Google, and found source code for an
implementation from Telchemy, which says its the reference on which the spec
was based.

I had to fix a number of things before this code would compile. I then found
a bug or two in the algorithms, and some more things I find unclear.

Is anyone using this code (or any G.1050 implementation), and interested in
a discussion about its finer points?

Regards,
Steve


_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www1.ietf.org/mailman/listinfo/avt


_______________________________________________
ippm mailing list
ippm at ietf.org 
https://www1.ietf.org/mailman/listinfo/ippm