Re: another laundry or shopping list

["Albert D. Cahalan" <acahalan@cs.uml.edu>]

David L. Nicol writes:
> Jordi Polo wrote:

>>> But using /proc to start work, people can focus on what features want
>>> there.
>
> Lets mount it, by default and for discussion purposes, at /proc/cluster

That means you need a mount point, which means more junk in /proc.
Forget about using /proc. You can have something similar to /proc,
but with separate code and a completely unrelated mount point.
Here is a Tru64 5.0 system:

$ ls -l /cluster/
total 3
drwxr-xr-x   3 root     system       512 Apr  5  2000 admin
drwxr-xr-x   3 root     system       512 Aug  4  2000 auth
drwxr-xr-x   3 root     system       512 Apr  5  2000 members
lrwxr-xr-x   1 root     system        14 Aug  4  2000 usr -> ../usr/cluster

That works for me.

> I like MOSIX's home-node paradigm.  A processes home node is where it 
> started, and the home node is responsible for keeping track of the
> process, wherever it goes, so it can send it signals and such, so when
> a process moves from one non-home node to another, it has to register
> this with its home node.
>
> Meaning in CPID terms, that a signal for a cpid would get redirected
> by the home node, which is obvious from the CPID.
>
> That's for process migration.

This is so... wrong. Now you have two nodes that must work OK,
increasing the chance of failure. You increase network traffic
and reduce performance.

There are two good ways to do process migration. The first way
is with SSI. The second is totally userspace. Doesn't Condor
support this? Obviously the userspace solution has limits, but
it ought to be the best performing.

> If all the different pieces operate independently
> 
> 	remote swap space

Remote swap space is already supported more or less. There might
still be some deadlocks. Shared swap space is more difficult.

Future directions seem to be toward using a filesystem-like swap
space, with anonymous files as backing store. This might help.

Peer-to-peer memory sharing would be cool.

> 	remote storage

Again, remote is easy. (the network block device) Shared is hard.

Linux-cluster: generic cluster infrastructure for Linux
Archive:       http://mail.nl.linux.org/linux-cluster/