Hello everybody, I was in silent mode in the list for the last 2 months,=20
and now I have a question, that really hope that someone could answer me.
First, let me explain the current situation.
We (Jan Westerholm, Mats Aspn=E4s and me) are biulding a cluster that=20
after have to make some races with another that are building at the=20
physics department, the point is that both teams have limited the funds=20
to 7.200 =80 and a HOWTo have to be written under FDL with all the=20
proccess of building and setting up it, we didn't decided about Beowulf=20
or OpenMosix yet, but is not that important because the hw and topology=20
will be same so ... (Mosix discarted due non-GPL compliance) and we=20
bought a node just to play with it and see the hw issues that we can=20
find. Well, righ now that node is working as OpenMosix cluster with my=20
desktop computer and looks great, processes migrating transparently and=20
all,I have no configuration questions (yet).
Today we find new questions that I will be extremely pleased if someone=20
can answer me (sorry if they are stupids) What happens if we ran=20
programs with braeks and debbuging from a front-end node, is that going=20
to overload that node because all the procceses have to come back when=20
they arrive to a break??? and second, what happens if we run a mpirun=20
program in an Openmosix cluster, we did guess about what can occurs what=20
we are not really sure.
The next step for us will be to buy 7 nodes and start building a bigger=20
prototype without my desktop computer on in (snif, it's nice to have two=20
processor just for me), and we will try these two problems on it. We=20
have calculated that maybe we can reach a number of 20 nodes with our=20
money. All these will be on the final document.
Thanks for all.
Sorry about my english, ussually is bad enougth but today I am exhausted=20
and shall we worst.