Re: tlug: Re: multi-processor linux configuration ?

["Manuel M. T. Chakravarty" <chak@example.com>]

"Andrew S. Howell" <andy@example.com> wrote,

> >>>>> "Manuel" == Manuel M T Chakravarty <chak@example.com> writes:
> 
>     Manuel> "Andrew S. Howell" <andy@example.com> wrote,
> 
[...]
>     Manuel> Broadcasting would definitely be worthwhile if you have a
>     Manuel> big number of machines.  But then, such a program would
>     Manuel> need a while to develop.  So how about the old parallel
>     Manuel> computing idea of realising broadcasts by a tree-like
>     Manuel> communication scheme.  Machine M1 rsyncs to M2, in the
>     Manuel> next step M1 rsyncs to M3 while M2 rsyncs to M4, in the
>     Manuel> next step M1 to M4 rsync four other machines, and so on.
>     Manuel> Easy to implement and rsync already uses time-stamps.
> 
> I have a bit of broadcast bias, as the core part of our business is
> doing reliable broadcasts. We have a daemon and some libs that make
> UDP reliable, even in the event of transient network problems.
> 
> One problem I see in doing a tree propagation is that if the machines
> are on the same local network, traffic is going to grow exponentially.

Right, but if somebody configures a big clusters of
workstations, which is to be used for parallel computing,
without a switch (as opposed to a hub), the whole set up is
pointless anyway (if the application is not really low in
communication).

Manuel
---------------------------------------------------------------
Next Nomikai: 20 November, 19:30 Tengu TokyoEkiMae 03-3275-3691
Next Meeting: 12 December, 12:30 Tokyo Station Yaesu central gate
---------------------------------------------------------------
Sponsor: PHT, makers of TurboLinux http://www.pht.co.jp