Re: [cf-dev] Problem while using Netem for benchmarking large-large post

[Abhijan Bhattacharyya <abhijan.bhattacharyya@xxxxxxx>]

Hi,
Thanks a lot for your response. We are
trying to bench-mark a typical large-large post scenario (the payload type
does not matter much; simple ASCII texts will do) under different network
bandwidth and loss situation. We are doing this as a background study which
would help in choosing the right candidate to build some solutions in future
which would require such large POSTs.
However, whenever we put the loss in
the system using 'Netem' or 'iptables' we see the odd behaviour. At times
the client code crashes as the POST handler, rather than going to the time-out-and-retransmit
loop in case of an emulated loss, gets a NULL response.  
Interestingly, since yesterday we started
experimenting the same scenario using CoAPthon (the python implementation
of CoAP). In this case the code does not crash, but occasionally stalls
indefinitely if there is a loss of control traffic.
Really no idea what is going on. With
emulated loss the system is becoming really unpredictable. At times the
exchange (about 4 blocks with 1024 bytes each) successfully happens with
several retransmits of the blocks. But at times it stalls after exchanging
one or two blocks.

We have created an isolated network
with two machines connected over a 3rd router (with 2 interfaces). The
loss is incorporated in the 3rd one. We have also experimented with back
to back connected client-server where the loss is applied at the server
NIC.

  
Regards
Abhijan Bhattacharyya
Associate Consultant
Scientist, Innovation Lab, Kolkata, India
Tata Consultancy Services
Mailto: abhijan.bhattacharyya@xxxxxxx
Website: http://www.tcs.com
____________________________________________
Experience certainty.        IT Services
                
       Business Solutions
                
       Consulting
____________________________________________




From:      
 "Hudalla Kai (INST/ESY1)"
<Kai.Hudalla@xxxxxxxxxxxx>
To:      
 "cf-dev@xxxxxxxxxxx"
<cf-dev@xxxxxxxxxxx>
Date:      
 10/20/2016 01:18 PM
Subject:    
   Re: [cf-dev]
Problem while using Netem for benchmarking large-large post in Californium
Sent by:    
   cf-dev-bounces@xxxxxxxxxxx

---

Hi Abhijan,

I have to admit that the blockwise support in Californium is not the best
maintained and understood area of functionality. We have even been thinking
about
re-factoring support for "transparent" blockwise transfers out
of Californium's
core stack and maybe into the CoapClient component only. However, I am
still
interested in your test case, so it would be helpful to know what you are
actually trying to do and, most importantly, what version of Californium
you are
using. If you think that what you experience is a bug, then you might also
consider creating an issue on GitHub [1].

[1] https.//github.com/eclipse/californium

-- 
Mit freundlichen Grüßen / Best regards

Kai Hudalla
Chief Software Architect

Bosch Software Innovations GmbH
Schöneberger Ufer 89-91
10785 Berlin
GERMANY
www.bosch-si.com

Registered office: Berlin, Register court: Amtsgericht Charlottenburg,
HRB 148411 B;
Executives: Dr.-Ing. Rainer Kallenbach, Michael Hahn

On Mo, 2016-10-17 at 12:39 +0530, Abhijan Bhattacharyya wrote:
> Dear list, 
> We are trying to do some bench marking of a typical large-large post
scenario
> under different bandwidth and losses. We are using Californium. However,
we are
> experiencing an apparently bizarre problem. 
> 
> We are running the client and server in two VMs and routing the packets
through
> a third VM. We are observing the exchanges in Wireshark at both client
and
> server ends. The exchanges happen as usual. However, whenever we are
> incorporating loss in the 3rd VM we see the problem. 
> 
> We are using the following command for incorporating loss in the 3rd
VM (this
> one is configured with 2 interfaces) 
> sudo tc qdisc add dev eth0 root netem loss <n>%
> 
> As soon as the loss is incorporated we see that after exchange of
one or two
> blocks (or some time not even a single one) the exchange stalls and
the client
> is getting Null pointer exception. Actually, the server is responding,
but the
> client is not able to receive the packets and the application layer,
rather
> than doing retransmits, is somehow crashing. Is Netem doing something
with the
> state machine? 
> If anyone has faced this problem or has any idea regarding this may
please
> respond. 
> 
> 
>   
> Regards
> Abhijan Bhattacharyya
> Associate Consultant
> Scientist, Innovation Lab, Kolkata, India
> Tata Consultancy Services
> Mailto: abhijan.bhattacharyya@xxxxxxx
> Website: http://www.tcs.com
> ____________________________________________
> Experience certainty.        IT Services
>                    
   Business Solutions
>                    
   Consulting
> ____________________________________________
> =====-----=====-----=====
> Notice: The information contained in this e-mail
> message and/or attachments to it may contain 
> confidential or privileged information. If you are 
> not the intended recipient, any dissemination, use, 
> review, distribution, printing or copying of the 
> information contained in this e-mail message 
> and/or attachments to it are strictly prohibited. If 
> you have received this communication in error, 
> please notify us by reply e-mail or telephone and 
> immediately and permanently delete the message 
> and any attachments. Thank you
> _______________________________________________
> cf-dev mailing list
> cf-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe
from
> this list, visit
> https://dev.eclipse.org/mailman/listinfo/cf-dev
_______________________________________________
cf-dev mailing list
cf-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/cf-dev