[Nfd-dev] Fwd: Loss Rate

[Lan Wang (lanwang)]

FYI, some information about the overlay link loss rate (measured with ping) over the testbed.  Loss rate is pretty low, e.g., 10 losses out of 100000 between UCLA and PKU at 100 pings per second (see below).

Lan

Begin forwarded message:

From: John DeHart <jdd at seas.wustl.edu<mailto:jdd at seas.wustl.edu>>
Subject: Re: Loss Rate
Date: August 19, 2014 3:30:23 PM CDT
To: "Lan Wang (lanwang)" <lanwang at memphis.edu<mailto:lanwang at memphis.edu>>, "marcsb at bellsouth.net<mailto:marcsb at bellsouth.net>" <marcsb at bellsouth.net<mailto:marcsb at bellsouth.net>>
Cc: "obaidasyed at gmail.com<mailto:obaidasyed at gmail.com>" <obaidasyed at gmail.com<mailto:obaidasyed at gmail.com>>


100,000 ping requests:

UCLA (spurs) to PKU (loss of 10 packets)

ndnops at spurs:~$ sudo ping -q -i 0.01 -c 100000 162.105.146.26
PING 162.105.146.26 (162.105.146.26) 56(84) bytes of data.

--- 162.105.146.26 ping statistics ---
100000 packets transmitted, 99990 received, 0% packet loss, time 1112399ms
rtt min/avg/max/mdev = 148.080/155.447/970.675/43.102 ms, pipe 96



TONGJI to CAIDA (loss of 8 packets):

ndnops at ndnops:~$ sudo ping -q -i 0.01 -c 100000 click.caida.org<http://click.caida.org>
PING click.caida.org<http://click.caida.org> (192.172.226.248) 56(84) bytes of data.

--- click.caida.org<http://click.caida.org> ping statistics ---
100000 packets transmitted, 99992 received, 0% packet loss, time 1371267ms
rtt min/avg/max/mdev = 181.647/202.676/251.866/12.110 ms, pipe 19

On 8/19/14, 2:37 PM, John DeHart wrote:

And here is a similar pair of results for 10000 pings trying at 100 requests per second.
Let me know if this level of result helps at all.

UCLA (spurs) to PKU:

ndnops at spurs:~$ sudo ping -q -i 0.01 -c 10000 162.105.146.26
PING 162.105.146.26 (162.105.146.26) 56(84) bytes of data.

--- 162.105.146.26 ping statistics ---
10000 packets transmitted, 10000 received, 0% packet loss, time 109146ms
rtt min/avg/max/mdev = 148.050/154.401/631.033/39.391 ms, pipe 63
ndnops at spurs:~$


TONGJI to CAIDA (1 loss out of 10000)

ndnops at ndnops:~$ sudo ping -q -i 0.01 -c 10000 click.caida.org<http://click.caida.org>
PING click.caida.org<http://click.caida.org> (192.172.226.248) 56(84) bytes of data.

--- click.caida.org<http://click.caida.org> ping statistics ---
10000 packets transmitted, 9999 received, 0% packet loss, time 136946ms
rtt min/avg/max/mdev = 181.710/193.677/210.785/5.104 ms, pipe 17
ndnops at ndnops:~$
On 8/19/14, 2:22 PM, John DeHart wrote:

Here are two initial tests doing 1000 ping requests,
using approximately 10 ping requests per second.

UCLA (spurs) to PKU:
ndnops at spurs:~$ sudo ping -q -i 0.1 -c 1000 162.105.146.26
PING 162.105.146.26 (162.105.146.26) 56(84) bytes of data.

--- 162.105.146.26 ping statistics ---
1000 packets transmitted, 1000 received, 0% packet loss, time 100958ms
rtt min/avg/max/mdev = 148.277/166.666/597.244/67.553 ms, pipe 6



TONGJI to CAIDA:
ndnops at ndnops:~$ sudo ping -q -i 0.1 -c 1000 click.caida.org<http://click.caida.org>
PING click.caida.org<http://click.caida.org> (192.172.226.248) 56(84) bytes of data.

--- click.caida.org<http://click.caida.org> ping statistics ---
1000 packets transmitted, 1000 received, 0% packet loss, time 106763ms
rtt min/avg/max/mdev = 181.729/224.260/270.358/19.935 ms, pipe 3



John

On 8/19/14, 2:15 PM, John DeHart wrote:

Lan,

The other issue I am encountering as I try some initial tests is
that not all sites seem to allow pings. For instance, Tongji,
one of the ones we would be most interested in, does not seem
to allow pings.

John

On 8/19/14, 2:09 PM, John DeHart wrote:

Lan,

How urgent is this? Are you trying to get some results before NDNComm?

John

On 8/19/14, 8:56 AM, Lan Wang (lanwang) wrote:
John,

For the international links, I am guessing 1 in 1000 or more.  Domestic ones may be lower (1 in 10000?)  These are overlay links, so they should have higher loss rate than a single physical link.

Lan



-------- Original message --------
From: John DeHart <jdd at seas.wustl.edu><mailto:jdd at seas.wustl.edu>
Date: 08/18/2014 7:36 PM (GMT-06:00)
To: "Lan Wang (lanwang)" <lanwang at memphis.edu><mailto:lanwang at memphis.edu>,marcsb at bellsouth.net<mailto:marcsb at bellsouth.net>
Cc: obaidasyed at gmail.com<mailto:obaidasyed at gmail.com>
Subject: Re: Loss Rate



Lan,

What are you expecting the loss rates to be? Are you expecting 1 in a million?

I would think we would have to run ping for a long time to get an accurate picture of the loss rate on a link.

John

On 8/18/14, 6:15 PM, Lan Wang (lanwang) wrote:
John,

Sorry for the lack of details.  Marc should have given you some background.  We are trying to emulate the testbed and losses are an important characteristic of the topology that can affect the behavior of the protocols and performance.   It is also good for testing the robustness of our design and implementation.  Right now we do not have a good idea of the loss rate on the testbed links.

We are mainly interested in losses from one nic to another nic.  I think running a continuous ping with one packet every second would be good enough.

Lan



-------- Original message --------
From: John DeHart <jdd at seas.wustl.edu><mailto:jdd at seas.wustl.edu>
Date: 08/18/2014 4:23 PM (GMT-06:00)
To: marcsb at bellsouth.net<mailto:marcsb at bellsouth.net>
Cc: obaidasyed at gmail.com<mailto:obaidasyed at gmail.com>,"Lan Wang (lanwang)" <lanwang at memphis.edu><mailto:lanwang at memphis.edu>
Subject: Re: Loss Rate



Marc,

Wow. That sounds like a simple question but I'm guessing it opens up a lot of possibilities.
Do you guys have something that you do on your local testbed already?

What is it you are really interested in?  Are you looking for losses from when
data leaves one NIC to when it arrives at the remote NIC
or are you looking for losses from sending process to receiving process?

John

On 8/18/14, 2:50 PM, Marc Badrian wrote:
Hey John,


Is there any way you could get the loss rate for each link/between each site for the current testbed topology?



Thanks,

- Marc








-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/nfd-dev/attachments/20150213/9b738496/attachment.html>