[erlang-questions] non-atomic nature of mnesia:dirty_update_counter

Chandru <>
Tue Mar 8 07:39:28 CET 2011


On 7 March 2011 16:16, Marcos Ortiz <> wrote:

> On Mon, 2011-03-07 at 14:39 +0000, Chandru wrote:
> > FWIW, we use mnesia in about 20 different systems all doing various stuff
> in
> > the network. Yes, mnesia doesn't handle partitioned networks very well,
> but
> > we've just learned to work around this. The 2G limit is not a limit at
> all -
> > just fragment your table. Our customer database of about 60 million
> entries
> > is partitioned into 128 fragments and it handles about 350 million
> lookups a
> > day, and the CPUs are rather bored on that machine.
>
> Chandru, can you share with us this knowledge in a article or a talk
> about this work with nmesia?
>
>
I'll try to write this up some time, but there are people such as Ulf who
have far more knowledge about mnesia than I do. I suggest you monitor his
work on mnesia closely :)

cheers
Chandru


More information about the erlang-questions mailing list