radzy@unet.UUCP (Tim Radzy) (03/03/89)
I need to use a distributed make system. We have about
200 SUN workstations, most of which will soon have local
disks with / and /usr (and thus, /tmp). We are running
medium-sized makes fairly frequently, and we would like
to be able to make use of all the aggregate CPU horsepower.
Before I go off and build something like this, does anyone
out there know of anything which will do what I want?
Of course, I'll be flexible about things. If someone out there
has something *remotely similar* to what I want, I'd really
appreciate hearing about it.
I'm thinking of something which has 2 programs: the user-level
program which I'll call "dmake", and a daemon which I'll call
"dmaked".
What I have in mind is that dmake would do the following:
1. parse the "makefile" and finds out what needs to be
done.
2. Send out a query to the other machines on the network,
and asks which machines are willing to do some work.
Dmaked would respond if it should (see below).
3. Send out messages to dmaked on each system, with the commands
and environment to run the particular task.
and dmaked would do the following
1. wait for a message from a dmake somewhere
2. If the dmake specified something that this machine can
do, and the load on this machine is light, then dmaked
should respond to the broadcast. Otherwise, it should
wait for another broadcast message.
3. when the dmake sends back a command, dmaked should run
that command and send back the status.
If you have something like this, please either respond to
this message (via mail or replies) or call me at (415) 780-5481.
I'll probably be out of the office, but PLEASE PLEASE PLEASE
leave a message.
-- Tim "radzy" Radzykewycz
The Incredible Radical Cabbage
{bacbell | pyramid | ames} !unet!radzy