Hello Martin I can get you access to a computer lab with about 10 computers that were used for a computer lab in wireless security. A re-install (as the students have changed the passwords) from image will give you a clean machine with a 2-3 months old Ubutu installation with access root if that is needed.
I will soon post another email to the list with sketches for an improved comparison. It is basically the same as previous work only looking at 2 bits at a time to speed things up. Martin Geisler wrote: > Tord Ingolf Reistad <[EMAIL PROTECTED]> writes: > > Hi Tord > >> I have one personal computer that I can run VIFF on here in Norway. >> >> I might try to get you access to a computer lab if you want, so you >> can run everything from Denmark and not worry about contacting me. All >> the students are studding for their exams so the labs are empty some >> time in January. > > That would be a great solution, so that I don't have to coordinate with > you to do my tests. An account with SSH access, 50 MiB of disk, and a > standard install of Python and GCC is all I need. > > But if you have the time, we could also schedule a benchmarking session. > It think it could be done in an hour -- less if you have already > installed Twisted and GMPY. But let me know if you're to busy right now. > >> I am only looking at the code and not understanding too much so I have >> still not tried to get VIFF working in practice. > > Okay, let me know if I can help! > >> When I talked to someone (more linux geek than me) about running MPC >> with computers far away, he suggested simulating the whole thing by >> introducing controlled delays in the network layer. But I like your >> idea better. > > For benchmarking I think it is best to run the programs on the real > Internet. But for testing I would love to have such a sub-system. It > would also make it possible to repeat and automate the tests in a > reliable way. > > I have a similar item on the TODO list already: > > * Create unit tests which randomizes network delay > The current unit test with LoopbackRuntime are strictly sequential and > does not capture the random delays possible with real network traffic. > > This is not so much to test the performance, but instead the robustness > of the network setup. Right now you have to start the players in order > 3-2-1 and if you kill one of them the port is probably left in a > TIME_WAIT state for 60 seconds. > > > > ------------------------------------------------------------------------ > > _______________________________________________ > viff-devel mailing list (http://viff.dk/) > [email protected] > http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk _______________________________________________ viff-devel mailing list (http://viff.dk/) [email protected] http://lists.viff.dk/listinfo.cgi/viff-devel-viff.dk
