[erlang-questions] Is mnesia fit to over TB data handling?

Tom Samplonius tom@REDACTED
Sat May 12 09:45:02 CEST 2007

----- "Tim Uckun" <timuckun@REDACTED> wrote:
> On 5/11/07, Ewan Higgs <ewan_higgs@REDACTED> wrote:
> > I think this study will be of interest to you:
> That was indeed interesting thanks.
> It seems to me they should have evaluated mysql cluster. It's feature
> set is closest to mnesia and it promises to be suitable for telecom
> level uptime.

  MySQL Cluster is not really a replicated database like Mnesia though.  In a MySQL Cluster all data nodes have to be tied together by high-speed connects.  Also as you add data nodes, query throughput drops, due to locking overhead.  That is unavoidable in a pure cluster though, as locks need to synchronously applied to all nodes, as as the number of data nodes increase, so does the locking overhead.

  Mnesia is rather different.  It can be synchronously updated, or not.  Supports "dirty" transactions if you need.  So it would scale to higher latency network links.


More information about the erlang-questions mailing list