For embedded systems without swap, overcommit is neccesary.
Imagine a firewall running off a floppy. It has a small application
proxy running on port 80 to forward WWW requests. Each instance of
the proxy handles 100 connections and then forks a new instance.
Each new instance gets a copy on write mapping of the same pages as
the original process, and unless you allow overcommit, you'll have to
reserve all this memory in case it will get written to.
So assuming a memory usage of 1 MB, (including the shared libraries
which are mapped into each instance) on a system with 8 MB of free
memory, you'll only be able to run 8 instances of this proxy.
In reality, most of this memory usage will be static mappings of
shared libraries and code, so by allowing overcommit, a small box such
as this will be able to run maybe a hundred instances or so instead.
/Christer
-- If it's tourist season, why can't we shoot them?- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/