[j-nsp] R: RE : multicast issue

[Alex Arseniev]

----- Original Message ----- 
From: <david.roy at orange.com>
To: "dim0sal" <dim0sal at hotmail.com>
Cc: <juniper-nsp at puck.nether.net>
Sent: Tuesday, July 16, 2013 8:49 PM
Subject: Re: [j-nsp] R: RE : multicast issue


> If you can't start/stop manually mcast streams you will never have stable 
> counters and can't detect packet loss.

This is not exactly true.
What You can do with constantly running streams is to have a FW filter with 
specific term matching on chosen stream IP headers+"count" action, on both 
headend (HE) and tailend (TE). Then activate/deactivate  this term as 
required. You have to have both HE and TE synced via common NTP source and 
perform commits at precise times ("commit at" if You do it manually).
Assuming headend and tailed are same router models and You have no network 
loss, You MAY get some minuscule counter difference between HE/TE because 
time spent by Routing Engine doing commit slightly varies. What would be 
more important is to monitor whether this difference changes with time which 
indicates packet loss.
HTH
Thanks
Alex