Hi All,

I've installed RSVP/ns and it runs well with examples in RSVP/ns document by 
Marc Greis..
I've ran ns~\tcl\ex\rsvp\rsvp_distinct.tcl successfully, and got the same data 
as Figure 5( Measurements for distinct reservations) in RSVP/ns document.

Then,I added trace-ralated codes in rsvp_distinct.tcl, to get delay,jitter,and 
loss for each flow. I found when a flow is successfully reserved,its loss 
become much lower than other BestEffort flows,so the receiver(node0) get more 
bits from reserved flow.But its delay,jitter become much higher than other 
BestEffort flows. e.g.When node0 have reserved 1k for node2 during 300s~900s, I 
get the following statistics:
 
SrcNode DstNode      delay(s)      jitter(s^2)     loss
2             0            1.350703      2.333099        0.326804
3             0            1.046185      0.000005        0.403614
4             0            1.046090      0.000004        0.392826

I've check my awk statistics script,there's no problem in it.And I've tried 
other periods(900s~1200s,1200s~1500s,etc.),all statistics are similar to above.

In my opinion,RSVP is expected to provide lower delay and jitter(or an 
upperbound) than other BestEffort flows,though only Controlled Load Service is 
supported in RSVP/ns.Am I right?

How to explain this phenomena? Is higher delay and jitter a cost for lower 
packet loss?

I would appreciate your help.
Thanks very much~

Best wishes,
michael.

Reply via email to