[erlang-questions] large scale deployments and netsplits

[Joel Reymont]

On Sep 16, 2009, at 9:02 PM, Scott Lystig Fritchie wrote:

> ... are the sizes of the Erlang VM processes big enough to run the  
> risk
> of paging virtual memory pages to disk?  If yes, then all bets are off
> for timely responses to network distribution communication, including
> liveness checks.

I took a look at one of the nodes that just split and beam was using 30%
of the memory and just 10% of the CPU. I don't think swapping is a  
problem.
It still dumbfounds me that nodes split when my own latency-measuring
process reports round-trips of a few milliseconds to just a few seconds.

I shudder at the thought of using Mnesia in this scenario! It used to be
that I thought of net splits as a remote possibility but at least with  
EC2
they are a solid reality.

---
fastest mac firefox!
http://wagerlabs.com