[ale] mosix clusters?

[Jeff Hubbs]

On Sun, 2002-06-23 at 21:04, Christopher Fowler wrote:
> My point is that some clusters like beowulf require an API to take
> advantage of them.  You have to write your code to use the cluster.

Right.  It's a different kind of clustering - more tightly coupled than
what Mosix, Turbolinux Cluster Server, or VMS do.  There's an API that
you use and you have to write your code just so.  It's for problems in
which the different processes ARE DEPENDENT on the others.  Beowulf
clusters also tend to get hurt when the machines aren't all the same,
whereas with Mosix, each node pulls what load it can pull.  
 
> Otherwise you have wasted your money.  I do not know much about a Mosix
> cluster.  So I'm not sure if it requires message passing or some other
> form of communications to get heavy computations done. 

Nah.  You can run all kinds of things - even different things at the
same time.  Mosix is basically a kernel patch and some config files.  

SETI is an
> example that would benefit from a Mosix cluster.

Not really.  SETI at Home work units take hours to days for to complete and
you'd have to create a submittal mechanism; SAH's default behavior is to
launch and execute new work units in series in the same PID.  In Mosix,
no new processes means no migration.  There is also the issue of whether
or not there is any benefit to running SAH more than once on the same
[1-CPU] machine; I assume not.  This implies that the best, simplest SAH
cluster is really not a cluster at all - just a machine farm, each one
running SAH independently.  That act will have the same effect as a
perfectly-load-balanced Mosix cluster running SAH.  

To revisit a bit, you'd need to know if, on a given 1-CPU machine, two
SAH WUs run at once is faster than, slower than, or as fast as two WUs
run end-to-end in the typical fashion.  If it's slower, then there is
absolutely nothing to be gained by running more SAH WUs at once than you
have CPUs.  My experience suggests that it would be slower, implying
that one instance per CPU is OPTIMAL; I say this because on a 2-CPU
machine, both CPUs get saturated, which tells me that memory I/O is not
an issue.  One instance per CPU is no different than what I do on every
machine I can get my hands on, Mosix-less.



> 
> 
> On Sun, 2002-06-23 at 18:49, Stephen Turner wrote:
> > i dont think it has an api, i believe i read that in the doc's, however
> > some clusters have the extra node(s)  as mirror images of the rest, so
> > they all contain the exact image hard disk, mosix says it doesnt have to
> > have the same hardware or size disks, thus i ask, does each node get
> > mirrord and all of them have the same hard disk image? or does all the
> > hard disks become one very large disk? and if so what happens to the data
> > when one node crashes? just gets lost?
> > 
> > __________________________________________________
> > Do You Yahoo!?
> > Yahoo! - Official partner of 2002 FIFA World Cup
> > http://fifaworldcup.yahoo.com
> > 
> > 
> 
> 
> 
> ---
> This message has been sent through the ALE general discussion list.
> See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
> sent to listmaster at ale dot org.
> 



---
This message has been sent through the ALE general discussion list.
See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
sent to listmaster at ale dot org.