[Spread-users] Simulating failure in SPREAD

Yair Amir yairamir at cnds.jhu.edu
Thu Nov 20 09:16:43 EST 2003


I am not aware of an easy way to simulate network delay unless you
are using something like Emulab www.emulab.net

However, there is a way to intentionally generate message
losses even though they are not lost. This may help you demonstrate
things. But to activate this you will have to change the code a bit:
In the file protocol.c, in the routine Prot_handle_bcast
look for the words "delete random" and uncomment the 3 lines
of code that are just below it. This will force Spread to randomly
get rid of messages.

Creating a partition with this modified daemon as messages are
flowing fast (e.g. using spflooder) will likely result in
the impact you are looking for.

    Hope this helps,

    :) Yair.    http://www.cs.jhu.edu/~yairamir

On Wednesday, November 19, 2003 7:18 PM
Michael Pilling Michael.Pilling at dsto.defence.gov.au wrote:

Michael> Dear All,
Michael> I wonder if you know of any simple way to simulate failure
Michael> in SPREAD? I want to do this so I can give a demo
Michael> to my superiors of the advantages of spread under failure.

Michael> I know one of the utilities allows you to create fake partitions
Michael> so that you can see it break into 2 subgroups and that works fine.

Michael> I'm wondering if there is a way to simulate delay between message
Michael> senders and receivers so we can force a partition while
Michael> the message is in transit.

Michael> Regards,

Michael> Michael Pilling

Michael> _______________________________________________
Michael> Spread-users mailing list
Michael> Spread-users at lists.spread.org
Michael> http://lists.spread.org/mailman/listinfo/spread-users

More information about the Spread-users mailing list