olson at mcs.anl.gov
Wed Jun 23 13:49:30 CDT 1999
>This is Kai Chen at the NLANR group at NCSA. I got your email message cc'd
>Speaking of network monitoring, you may have known the two current network
>1) NLANR: AMP (Active Measurements Project, http://amp.nlanr.net/)
> Things being measured: round-trip delay, packet loss, on-demand
>2) Advanced.org: Surveyor (http://www.advanced.org/surveyor/)
> Things being measured: one-way delay, packet loss.
>In each of these projects, a set of machines at participating HPC sites are
>deployed to run the testing programs regularly, and then send the results
>to a central database for post-processing and viewing. You may look at the
>web pages for the performacne data they are collecting.
Cool -- I'd not run across those before.
>>From the message below, I am not sure exactly what you are looking for in
>terms of monitoring the links. Which tools are you using for viewing the
>links to Atlanta? If you are looking for exactly the metrics AMP and
>Surveyor are collecting, I would suggest installing a monitor on the sites
>you are interested in. If you have special needs for monitoring other
>metrics, and would like to deploy this new system on your own nodes, we may
>be able to customized some tools for you.
This is an approach we should look at. We need to do some work in figuring
out exactly what it is we need to know about the network, but the basic
latency/jitter stats are a good starting point. The number we'd really like
is available bandwidth, but that's probably hard :-).
Off the top of my head, the statistics that I find myself wanting are
per-site, per-stream bandwidth/loss/latency/jitter statistics for the media
streams. This is information that can be gleaned from the RTP/RTCP headers
and sender/receiver reports. rtpmon provides some of this information, but
not all. It would also be nice to be able to identify the critical paths
thru the network between sites, and the amount of bandwidth we are using on
More information about the ag-tech