[nets] [Nets-seminars] Seminar on Friday

[Raul Landa]

Hello all,

In the questions session after the talk, I mentioned a new mechanism  
to create virtual networks in user space that uses the actual  
protocol stacks and scales to up to at least  1000 hosts and 33  
switches in a single laptop by using light virtualization.

This is the reference:

http://conferences.sigcomm.org/hotnets/2010/papers/a19-lantz.pdf

Of particular interest to the discussion at the time are  Tables 2, 3  
and 1, along with Section 5, where the authors discuss the  
limitations of their scheme.

Cheers
Raul



On 7 Nov 2011, at 5:05PM, Richard G. Clegg wrote:

> This Friday's talk is in GS/302 and will take place at 16:00.
>
> Richard Clegg and Stuart Clayman -- Experiments with virtual routers
>
> Validating research ideas for new protocols or architectures  
> requiring large topologies is difficult. Simulation results can be  
> doubtful and referees often raise queries about the quality of  
> simulation used. On the other hand, testbed work can often limit  
> the size of network investigated. Even with system virtual machines  
> such as Xen or VMWare, the number of nodes tested can be quite  
> small. In this work a lightweight router running in a java virtual  
> machine (JVM) has been developed. This is a fully functioning  
> router which can be used to test research ideas. Existing software  
> can be adapted to the virtual router socket API and hence the  
> software can be tested on a realistic large-scale architecture and  
> in an environment where routing and packet delivery can be  
> unreliable. Even on a modest sized testbed with only nine machines,  
> topologies with up to 720 routers have been tested. This talk  
> describes the virtual router and the testbed which has been  
> developed and the software which has been ported to use it. A  
> motivating example of research into placement and selection of  
> management and monitoring nodes is described in detail to show how  
> the platform has been successful in testing algorithms.
>
> -- 
> Richard G. Clegg,
> Dept of Elec. Eng.,
> University College London
> http://www.richardclegg.org/
>
>
> _______________________________________________
> Nets-seminars mailing list
> Nets-seminars at cs.ucl.ac.uk
> http://oakham.cs.ucl.ac.uk/mailman/listinfo/nets-seminars
>
> _______________________________________________
> nets mailing list
> nets at cs.ucl.ac.uk
> http://oakham.cs.ucl.ac.uk/mailman/listinfo/nets