[erlang-questions] "Erlang, the next Java"
Thu Aug 16 04:07:02 CEST 2007
The problem is that 32 CPU systems are expensive. You could afford
more CPU time if you did a cluster of 2 or 4 proc boxes. In fact, it
is the dominant strategy to create large computing centers.
If I had to do it, I would write my own custom distribution protocol
in pure erlang.
Sent from my iPhone
On Aug 15, 2007, at 12:47 PM, Brian McCallister <brianm@REDACTED>
> On Aug 11, 2007, at 2:54 PM, Ulf Wiger wrote:
>> If you want just 1000 CPUs, why not connect 32 nodes, each using
>> 32 cores? Then you don't even have to resort to trickery. ;-)
> If, say, you are writing a management agent to run on tens or
> hundreds of thousands of machines so you want the distribution and
> process monitoring more than processing power...
> Being able to use Erlang's default distribution model is a major
> benefit. You cannot really do a flat topology, but, frankly, Erlang
> is kinda beautiful for this kind of thing.
> Speaking of global (not quotes) -- do you know the mutex algorithm it
> uses for global locks? Is it possible to use global for a subset of
> nodes (based on number you are willing to tolerate failing or the
> number of monitored child nodes a single can support (ie, 50 servers
> coordinating 1000 children each gives you 50k in a very shallow tree)?
> erlang-questions mailing list
More information about the erlang-questions