[Mini-NDN] result of a topology

Mon Nov 21 20:21:32 PST 2016

Yes, we fail a node in the experiment for some time period.

The red line that has a gap is for single path routing/forwarding.  You need to set the multipath factor to 1 (number of faces in the FIB for each prefix) in the NLSR configuration to achieve this effect.  If I remember correctly, when the number is 0, the router will put all the faces in the FIB entry (ranked by their path cost).

We have an updated version of the paper as an NDN technical report: https://named-data.net/wp-content/uploads/2016/01/ndn-0037-1-nlsr.pdf.  It doesn’t have this specific figure though.

Lan

On Nov 21, 2016, at 4:01 PM, sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>> wrote:

thank you very much

On Tue, Nov 22, 2016 at 1:29 AM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

Yes it will happen automatically. I think in the paper the node was down for 2 minutes then brought up.

Ashlesh

________________________________
From: sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Monday, November 21, 2016 3:57:39 PM

To: Ashlesh Gawande (agawande)
Cc: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: Re: [Mini-NDN] result of a topology

so you mean that it will automatically happen and I shouldn't consider anything for it in my experiment?

On Tue, Nov 22, 2016 at 1:18 AM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

Oh I see, convergence for the best path will happen automatically once the node recovers (i.e. once NLSR comes back up on the failed node).

Ashlesh

________________________________
From: sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Monday, November 21, 2016 3:46:05 PM

To: Ashlesh Gawande (agawande)
Cc: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: Re: [Mini-NDN] result of a topology

then how can that gap be explained?
I mean then why that convergence gap is considered in the paper's figure?

On Tue, Nov 22, 2016 at 1:12 AM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

So when the node is failed, NFD should automatically use the next best path installed by NLSR. Convergence is already checked before running the experiment: https://github.com/named-data/mini-ndn/blob/master/ndn/experiments/experiment.py#L82

If it does not, then maybe the max-faces-per-prefix is set to 3 by default (https://github.com/named-data/NLSR/blob/master/nlsr.conf#L115) .You can use the option --faces 0 in Mini-NDN to use all the available paths and then it should use the next best path. This probably won't happen with this small topology.

Ashlesh

________________________________
From: sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Monday, November 21, 2016 3:36:44 PM

To: Ashlesh Gawande (agawande)
Cc: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: Re: [Mini-NDN] result of a topology

ok thank you very much
I also have another question:
as you can see, the red marks in the paper's figure have a gap of about 60 seconds after that node4 fails,
how can I produce this gap or in another words cause convergence after node's failure in my experiment, I know how to put the node off and how to get in back on but I don't know how to cause convergence in my experiment...?

On Tue, Nov 22, 2016 at 12:40 AM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

You can try 20 or 30 ms and see if you get better results?

Ashlesh

________________________________
From: sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Monday, November 21, 2016 3:03:47 PM

To: Ashlesh Gawande (agawande)
Cc: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: Re: [Mini-NDN] result of a topology

what would a reasonable delay be?

On Tue, Nov 22, 2016 at 12:11 AM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

Why is the delay zero, I think non-zero delay might give smoother result.

Do you want to send 99999999 pings in 200 seconds?

ndnping allows to send at most 1000 pings/second, so in 200 seconds you can send 200,000 pings at best by using the option "-i 1" here:

https://github.com/named-data/mini-ndn/blob/master/ndn/experiments/experiment.py#L92

You can modify ndnping if you want to have more pings than that.

(This will be heavy on the CPU).

Ashlesh

________________________________
From: sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Sunday, November 20, 2016 9:54:22 AM

To: Ashlesh Gawande (agawande)
Cc: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: Re: [Mini-NDN] result of a topology

I changed the cpu of the machine and the figure became like:
<photo_2016-11-20_19-21-00.jpg>
as you can see the swap is ok
now I've got a question about the number of pings?
how can I write the experiment to have the maximum number of pings in 200 seconds.
my current experiment is :
from ndn.experiments.experiment import Experiment
from ndn.nlsr import Nlsr

import time

class pingallreg(Experiment):

    def __init__(self, args):
        args["nPings"] = 99999999
        Experiment.__init__(self, args)

    def run(self):
        self.startPctPings()
        time.sleep(200)

Experiment.register("wofail", pingallreg)

which the number 99999999 is to have as many pings as possible,
how can I get the maximum pings possible?

On Sun, Nov 20, 2016 at 5:34 PM, sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>> wrote:
the delay is zero
the topology file is:
[nodes]
node4: _ hyperbolic-state=off radius=0.0 angle=0.0 network=/ndn router=/%C1.Router/cs/node4 site=/edu/site nlsr-log-level=NONE max-faces-per-prefix=0 nfd-log-level=NONE
node1: _ hyperbolic-state=off radius=0.0 angle=0.0 network=/ndn router=/%C1.Router/cs/node1 site=/edu/site nlsr-log-level=NONE max-faces-per-prefix=0 nfd-log-level=NONE
node5: _ hyperbolic-state=off radius=0.0 angle=0.0 network=/ndn router=/%C1.Router/cs/node5 site=/edu/site nlsr-log-level=NONE max-faces-per-prefix=0 nfd-log-level=NONE
node3: _ hyperbolic-state=off radius=0.0 angle=0.0 network=/ndn router=/%C1.Router/cs/node3 site=/edu/site nlsr-log-level=NONE max-faces-per-prefix=0 nfd-log-level=NONE
node2: _ hyperbolic-state=off radius=0.0 angle=0.0 network=/ndn router=/%C1.Router/cs/node2 site=/edu/site nlsr-log-level=NONE max-faces-per-prefix=0 nfd-log-level=NONE
node6: _ hyperbolic-state=off radius=0.0 angle=0.0 network=/ndn router=/%C1.Router/cs/node6 site=/edu/site nlsr-log-level=NONE max-faces-per-prefix=0 nfd-log-level=NONE
[switches]
[links]
node1:node2
node1:node3
node3:node5
node5:node6
node6:node4
node4:node2
node4:node3
node3:node6

On Sat, Nov 19, 2016 at 5:27 PM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

No, the figure of the paper should be possible to achieve. The CPU is the bottleneck because one has to run 6 NFD, 6 NLSR instances on a single machine. I have run a 10 node pingall experiment on a VBox VM with 2.5GB RAM and 2 CPUs without so much variation.

To turn off the logs:

Change DEBUG to NONE here:

https://github.com/named-data/mini-ndn/blob/master/ndn/nlsr.py#L123

and make sure that this also set to NONE:

https://github.com/named-data/mini-ndn/blob/master/ndn/nfd.py#L34

Then reinstall Mini-NDN, sudo ./install.sh -i

Also make sure that in your topology file all the log levels are set to NONE if there is such an option after the nodes.

Then try to run the experiment again (If possible turn off the GUI in VM, i.e. login in to command line and run the experiment from there - just to free up some more CPU).

In your graph the y-axis is the RTT in ms? What are the delays you have set on the links of the topology?

(If possible, can you share the topology file?)

Ashlesh

________________________________
From: sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Saturday, November 19, 2016 12:48:23 AM
To: Ashlesh Gawande (agawande)
Cc: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: Re: [Mini-NDN] result of a topology

Hello,
no there aren't any timeouts and what you see there is not a zero it is 10^0 which is equal to 1.
I'm running my linux on a vmware virtual machine with the configurations bellow:
<image.png>
so you're saying that this results which I've got in my figure is reasonable for the topology and it shouldn't exactly be like the figure in the paper?
Thank you

On Fri, Nov 18, 2016 at 4:53 PM, Ashlesh Gawande (agawande) <agawande at memphis.edu<mailto:agawande at memphis.edu>> wrote:

Have you verified that the ping data is correctly depicted on your figure?

Are there any timeouts in there? Because I see some zeros.

RTT can be have variation in Mini-NDN because of multiple NFDs are running on the same machine.

What is your current machine configuration?

Usually for such a small topology it should not be a problem.

One thing you can try is that if you are collecting NFD, NLSR, or any other logs, turn them off.

That is turn off all the logs except the ping data.

Other thing to try is to use a machine with more CPU power.

Ashlesh

________________________________
From: Mini-NDN <mini-ndn-bounces at lists.cs.ucla.edu<mailto:mini-ndn-bounces at lists.cs.ucla.edu>> on behalf of sepehr abdous <sepehrabdous1375 at gmail.com<mailto:sepehrabdous1375 at gmail.com>>
Sent: Friday, November 18, 2016 6:48:35 AM
To: mini-ndn at lists.cs.ucla.edu<mailto:mini-ndn at lists.cs.ucla.edu>
Subject: [Mini-NDN] result of a topology

Hello all,
I'm new to mini-ndn and I'm trying to implement the following topology:
<Untitled.png>
so I use minindnedit to implement it with the following information:
<Untitled.png>
and based on the paper which represents this topology it is expected to get this figure from it's information:
<Untitled.png>
which the blue "+" represents the ping-data in the case that no node fails, but when I try to implement it using pingall experiment my figure will come out like this:
<image.png>
as you can see there is more swing in the blue '+'s than it is expected, any idea why it is happening?
thank you

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran

--
Sepehr Abdous
Faculty: abdous at ce.sharif.edu<mailto:abdous at ce.sharif.edu>
Yahoo mail: sepehrabdous at yahoo.com<mailto:sepehrabdous at yahoo.com>
Cellphone:+989194030591
School of computer engineering
Sharif University of Technology
Tehran, Iran
_______________________________________________
Mini-NDN mailing list
Mini-NDN at lists.cs.ucla.edu<mailto:Mini-NDN at lists.cs.ucla.edu>
http://www.lists.cs.ucla.edu/mailman/listinfo/mini-ndn

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/mini-ndn/attachments/20161122/2c29e239/attachment-0001.html>