Re: [nistnet] Packets queued up indefinetly

Thu, 27 Jul 2006 23:23:23 -0700

Hi,
Thank You for your replies.

'rmmod nistnet' didnt help me because, packets are getting queued up every time I start, its not that packets are queued up .
I am not at all sending packets faster than the NistNet emulated bandwidth....do I have to set bandwidth and drop outs inorder to emulate delay ?

Or else, can someone please tell me the right kernel version on which NIST Net runs without any issues. Is there any hardware dependency for NIST Net to work ?

Thank You.
--
Naresh
http://students.iiit.net/~satyanaresh/
--

On 7/28/06, Erwin Himawan <[EMAIL PROTECTED]> wrote:
Try to remove the module -> rmmod nistnet and after that load nistnet module again.
 
I experienced this problem if I have been running NistNet for days, and after that Linux just quit forwarding any packet at all.  I have to reboot the machine to get the routing function into normal operation.
 
Erwin
----- Original Message -----
From: David Oltman
Sent: Thursday, July 27, 2006 4:11 PM
Subject: Re: [nistnet] Packets queued up indefinetly

I've seen a problem like this on our NIST Net "router" (a PC configured with two NICs).  Our "router" works fine if we limit bandwidth just once after booting the machine.  However, if we try to change the bandwidth limiting factor more than once during a single session, NIST Net will queue all packets on the connection and all transmission will stop.  Rebooting the machine fixes the problem.  We may be doing something wrong, but I'm at a loss to figure what that might be.

Dave Oltman
SAIC, Inc.

Erwin Himawan wrote:
Did you sending packet faster than the NistNet emulated bandwidth?  If you did this, you have to set NistNet Drop mechanism as well.  I believe NistNet uses DRD algorithm to drop the packet due to congestion.  If you do not configure DRD, you will experience increasing delay as time progressing.
 
When you are introducing delay; you are indirectly affecting the bandwidth.  For example, you have 10mbps bandwidth, which mean you inout 10mb, after 1 second you got 10mb; so your bw is 10mbps.  Assuming you are putting 1 second delay, the affect is you are reducing the bandwidth by half; i.e after two seconds you got the 10mb, which means your bw is 5 mbps.  As you are sending at constant data rate 10mbps to 5mbps bw, your emulated network experiencing congestion.  Without congestion avoidance configured at NistNet, NistNet will buffer the input data rate infinetely and the result is you observe increasing delay as time is progressing.  I believe if you are sending rate is less than 5mbps, you should not see constant delay instead of increasing delay.
 
In order to pretty much emulate the network you are emulating, you need to find out what your network will do if the network experience congestion?  Does the network drop the packet? What algorithm the network choose to drop the packet? Tail, Front or Random?  What is the threshold when the network starts to trigger the congestion avoidance mechanism?
 
Hopefully, I understand your question correctly and provide the answer you are looking into.
 
Erwin
 
On 7/27/06, VVS.Naresh < [EMAIL PROTECTED]> wrote:
Hi,
 
I installed NIST Net on my Fedora Core-4 2.6.11 kernel and the problem I am facing is
all the packets are getting queued up on NIST Net machine(say System:A) which has
two NICs and is a router for two different domains. Packets are getting queued up if and
only if I start NIST Net with some emulation factor( I tried delaying the packets).
 
In my /proc/interrupts, I see fast_rtc module coming up whenever I start nistnet, but the
thing is it is not giving any interrupts(I see 0 in place of 8192Hz or whatever).
 
Can someone help me regarding this problem....
 
thank you,
regards,
VVS.Naresh
--

_______________________________________________
nistnet mailing list
nistnet@antd.nist.gov
http://www-x.antd.nist.gov/mailman/listinfo/nistnet




_______________________________________________
nistnet mailing list

nistnet@antd.nist.gov

http://www-x.antd.nist.gov/mailman/listinfo/nistnet

_______________________________________________
nistnet mailing list
nistnet@antd.nist.gov
http://www-x.antd.nist.gov/mailman/listinfo/nistnet





--

--
VVS.Naresh
Technical Member - VoIP,
HelloSoft India Pvt. Ltd.

MailTo: [EMAIL PROTECTED]
On Net: www.hellosoft.com
_______________________________________________
nistnet mailing list
nistnet@antd.nist.gov
http://www-x.antd.nist.gov/mailman/listinfo/nistnet

Reply via email to