Hardware for EROS?

[David L. Nicol]

Kragen Sitaker wrote:

> By the way, I'm interested to hear about your distributed-computing
> scenarios that require stable commitment.

I've been thinking about PC clustering using university lab machines;
a risk, of course, is that a lab machine will have its power cycled
without proper guest process expulsion procedures (mosix load balnacing)

The solution I plan on implementing, after gathering sufficent round
tuits, may involve duplicating "farmed out" work, or at least keeping
a back-up of the remote process image in synchronization with updates
to its dirty file system pages, possibly only during "danger" hours
when the labs may have people in them.

This makes for a large-grained solution, to catastrophic node failure,
which is a large-grained problem.

It may be easier to just extend the cluster into the labs during
hours when they are empty.

I too am interested to hear about Mark Miller's distributed-computing
scenarios that require stable commitment.  I wonder what granularity
he is working at.



_____________________________________________________________________
                             David Nicol 816.235.1187 nicold@umkc.edu
"Nobody wants a hammer with racing stripes and a horn." - Greg Knauss