[erlang-questions] very large distributed implementation?

Mojito Sorbet <>
Tue Aug 17 04:06:48 CEST 2010


Another issue is the preference for Erlang to connect all nodes in a
fully connected mesh of TCP links.   That is N-1 links per node.  This
can be disabled, and result  in tree connections instead of mesh
connections, but depending on the problem being solved maybe only
occasional connectivity is required.

On Mon, 2010-08-16 at 18:05 -0700, Michael Truog wrote:

> Any Erlang code that works on 100s or 1000s of nodes is unlikely to use
> the distributed Erlang mode because of the heartbeat message that is
> used every minute (i.e., so local network backbones that are not
> expected to go down).  That means that such Erlang code is likely to
> utilize tcp, udp, etc. with custom code that manages consistency.




More information about the erlang-questions mailing list