Re: Hurd on a cluster computer

[Olaf Buddenhagen]

Hi,

On Tue, Jul 26, 2016 at 01:42:16PM -1000, Brent W. Baccala wrote:

> I've been wondering about using Hurd on a cluster computer;
[...]
> If so, then the first step would be to modify Mach,
[...]

Pretty much everything you describe has actually been considered in the
original design of Mach, as it was meant to be the successor for certain
projects (Amoeba I think?) that were entirely about this topic.

However, as far as I can tell this has never been actually fully
implemented in Mach; and we dropped the remnants of the disfunctional
code for network IPC from our Mach variant several years ago.

While I'm not an expert on this, from what I have seen and heard, the
world has pretty much given up on this kind of transparent clustering,
as for pretty much all workloads that require such massive scaling for
performance / resource demands (large network servers, scientific
computing, render farms etc.), it's more effective and easier to manage
when partitioning the work explicitly using various kinds of specialised
protocols.

-antrik-